Novel Insight into the Formation of Odour—Active Compounds in Sea Buckthorn Wine and Distilled Liquor Based on GC–MS and E–Nose Analysis

Sea buckthorn wine (SW) and distilled liquor (DL) are fruit wines with beneficial health effects. However, their unpleasant flavour limits their development and widespread acceptance. Therefore, it is necessary to analyse their flavour composition and changes. In this study, differential metabolites of sea buckthorn DL during processing were analysed, and the relationships between E–nose sensor values and key volatile organic compounds (VOCs) were established. The results show that 133 VOCs were identified, with 22 aroma–contributing components. Fermentation significantly increased the content of VOCs, especially esters. A total of seven and 51 VOCs were significantly upregulated after fermentation and distillation, respectively. Meanwhile, seven sensors were positively correlated with the increased level of alcohols and esters, and reflected the increasing trends of 10 key VOCs.


Introduction
Sea buckthorn (Hippophae rhamnoides), classified as a medicine food homology (MFH) plant, contains large amounts of vitamin C, carotene, flavones, linoleic acid and other bioactive substances [1,2]. For example, the vitamin C content of sea buckthorn juice (SJ) reaches 1000 mg/100 g, which is two to three times higher than that of kiwifruit (Actinidia deliciosa) [3]. Sea buckthorn has attracted a great deal of attention in recent years due to the functions of preventing blood stasis, strengthening the spleen and stomach and even fighting against cancer [4]. However, the sour taste existing in sea buckthorn has resulted in its application being confined to the products of juice, yogurt and oil.
Sea buckthorn wine is obtained by fermentation of SJ with yeast or koji after acid reduction and sugar adjustment. Negi [5] developed sea buckthorn wine with significant antioxidant activity and higher levels of flavonoids, quercetin, waxberry and rutin compared to Sauvignon salad wine. The report also showed that it had a protective effect against the oxidative stress response induced by furone and high-cholesterol diet-induced hypercholesterolemia in male mice of the LACA strain. However, the industrial development of this potentially beneficial product is limited because of its unpleasant flavour. Therefore, it is necessary to analyse the flavour of SW and sea buckthorn distilled liquor (DL).
Many methods and instruments are available for the determination of volatile organic compounds (VOCs) in food, including gas chromatography (GC), gas chromatographymass spectrometry (GC-MS), electronic nose (E-nose), electronic tongue (E-tongue), etc. VOCs in food have complex compositions and uneven distributions, making it difficult to elucidate flavour information using a single analytical technique [6,7]. GC-MS is one of stability. The samples were inoculated with 0.2% (w/v) Saccharomyces cerevisiae (Angel 97 Yeast Co. Ltd., Hubei, China) that had been activated in 2% glucose solution at 37 °C 98 (water bath) for 30 min, and then underwent fermentation at 28 °C for 7 days (the CO2 99 produced by fermentation can be discharged into the water with the hose) to obtain SW 100 (SW1-SW3) with an alcohol content of 13 ± 2% (v/v). Finally, 1L SW was distilled, pro-101 ducing 500 mL distillate for the first time (GG-17 all-glass distiller, Guangzhou Diangrui 102 Glass Experimental Instrument Co., Ltd., Guangzhou, China), and 250 mL distillate liq-103 uor (DL1-DL3) with an alcohol content of 40 ± 2% (v/v) was obtained by second distilla-104 tion. The whole research process is shown in Figure 1. Based on the method of Xu et al. [20], 1 mL samples were obtained and made up to 2 110 mL with distilled water (diluted samples). To the tubes containing 2 mL of diluted sam-111 ple, we successively added 2 mL of 6 mmol/L FeSO4 solution and 2 mL of 6 mmol/L H2O2 112 solution, followed by shaking. The mixture was allowed to stand for 10 min at room 113 temperature. Then, 2 mL of 6 mmol/L salicylic acid was added and allowed to stand for 114 30 min at room temperature. Each sample was determined three times in parallel. 115 Ascorbic acid was used as a positive control. The OH· radical scavenging rate was de-116 termined using the following formula: where Ai is the absorbance of the sample at 510 nm, Aj is the absorbance measured after 118 replacement of H2O2 with distilled water, and A0 is the absorbance measured in the blank 119 control group with distilled water instead of the sample.

OH· Radical Scavenging Rate
Based on the method of Xu et al. [20], 1 mL samples were obtained and made up to 2 mL with distilled water (diluted samples). To the tubes containing 2 mL of diluted sample, we successively added 2 mL of 6 mmol/L FeSO 4 solution and 2 mL of 6 mmol/L H 2 O 2 solution, followed by shaking. The mixture was allowed to stand for 10 min at room temperature. Then, 2 mL of 6 mmol/L salicylic acid was added and allowed to stand for 30 min at room temperature. Each sample was determined three times in parallel. Ascorbic acid was used as a positive control. The OH· radical scavenging rate was determined using the following formula: where A i is the absorbance of the sample at 510 nm, A j is the absorbance measured after replacement of H 2 O 2 with distilled water, and A 0 is the absorbance measured in the blank control group with distilled water instead of the sample.

DPPH Clearance Rate
According to the method of Liu, Ooi and Chang [21], 2 mL of 0.04 mg/mL 2,2diphenyl-1-picrylhydrazyl (DPPH) solution was added to 2 mL of diluted sample, mixed, and allowed to react for 20 min. Each sample was determined three times in parallel. Ascorbic acid was used as a positive control. The DPPH clearance rate was determined using the following formula: where A i is the absorbance of the supernatant at 517 nm, A j is the absorbance measured after replacement of DPPH solution with anhydrous ethanol, and A 0 is the absorbance of the reference consisting of 2 mL 0.04 mg/mL DPPH and 2 mL of anhydrous ethanol reactant.

Reducing Activity
According to the method of Kaewnarin et al. [22], 1 mL samples were taken and made up to 2.5 mL with distilled water, followed by the addition of 2.5 mL of 0.2 mol/L phosphoric acid buffer (pH 6.6) and 2.5 mL of 1% potassium ferric cyanide solution. After incubation at 50 • C in a water bath for 20 min, the samples were rapidly cooled. Then, 2.5 mL of 10% trichloroacetic acid was added and centrifuged at 3000 rpm for 10 min. After centrifugation, 5 mL of the supernatant was taken, and 4 mL of distilled water and 1 mL of 0.1% ferric chloride solution were added and mixed by oscillation. The absorbance at 700 nm was measured after 10 min. Ascorbic acid was used as a positive control. Each sample was determined three times in parallel.

GC-MS Detection
According to the method of Xia et al. [23], with minor modifications, 0.5 g of NaCl and 100 ppm 3-octanol (internal standard substance) was added to 10 mL of sample solution in 20 mL sealed glass vials. The samples were extracted at 40 • C for 40 min with 50/30 µm divinylbenzene/carboxen/polydimethylsiloxane (DVB/CAR/PDMS) fibre (Supelco, Bellefonte, PA, USA). Flavour compounds were detected by GC-MS (5975 mass spectrometer coupled to a 7890A gas chromatograph; Agilent, Santa Clara, CA, USA). A DB-INNOWax column (60 m × 0.25 mm ID and 0.25 µm film thickness) was used for separation. The temperatures of the injector, electron ionisation source, quadrupole chamber and transfer lines were 250 • C, 230 • C and 250 • C, respectively. The initial temperature was 50 • C for 3 min, which was increased to 80 • C at a rate of 3 • C/min. The temperature was further increased to 230 • C at 5 • C/min and maintained at 230 • C for 6 min. The carrier gas had a flow rate of 1.0 mL/min. Samples were injected in splitless mode. A mass range of 50-550 m/z was recorded at a rate of one scan per second. Each sample was determined three times in parallel.
Data were analysed using GC-MS software (Agilent, CA, USA) and identified based on the NIST 2017 database. Only compounds with a matching degree ≥80 were retrieved and recorded. The NIST database was used to automatically retrieve mass spectral data of each component for qualitative analysis. A semiquantitative method was used to determine relative contents by calculating the ratio of the internal standard substance (3-octanol) to the peak area of each component. Principal component analysis was used to analyse the differences in the flavour of the compounds among the three types of samples (SJ, SW and DL). Odour activity value (OAV) was obtained by the ratio of the concentration of the substance to the threshold value.

E-Nose Measurement
In accordance with the method of Cui et al. [24], E-nose (PEN3, German AIRSENCE company, Schwerin, German) was applied with a detection time of 120 s, cleaning time of 80 s, pre-injection time of 5 s, injection flow rate of 400 mL/min, and carrier gas (clean air) flow rate of 400 mL/min. At the beginning of measurement, the sensor fluctuated with time and began to flatten after 110 s. The sample was diluted 1:40, and 10 mL was then added to a test tube (20 mL), which was subsequently sealed with plastic wrap. After 30 min in a water bath at 50 • C, headspace gas of the samples was analysed by E-nose. Data at 114, 115 and 116 s were used to calculate the average for analysis. Each sample was determined three times in parallel. The 10 sensor values corresponding to nine samples (S1-SJ3, SW1-SW3, DL1-DL3) were used for stacking bar chart (Origin 9.0 software, OriginLab Co., Northampton, NC, USA). Principal component analysis was used to analyse the differences of the overall odour among the three groups of samples (SIMACA-P software, Umetrics Co., Malmo, Sweden). Furthermore, the correlation analysis (R software, The University of Auckland, Auckland, New Zealand) between the response value of different sensors of electronic nose and the content of flavour substances in GC-MS was carried out to establish the relationships between E-nose data and key VOCs to provide a basis for the rapid detection and flavour optimisation of SW.

Statistical Analysis
SPSS17.0 software was used for variance analysis. A value of p < 0.05 represented significant difference, p < 0.01 represented extremely significant difference, and the experimental results were expressed as the mean ± standard deviation. Each index was repeated three times. Principal component analysis was performed using SIMCA-P software, line charts and bar charts were constructed in Origin 9, and R language was used for the generation of Venn diagrams, volcano diagrams, and correlation heat maps.

Physicochemical Characteristics of Sea Buckthorn Wine
The physicochemical indices and antioxidant activities of SJ, SW and DL are shown in Figure 2. The amount of soluble sugar in SJ was significantly decreased after fermentation, but no significant difference was observed after distillation ( Figure 2A). The pH of SJ remained between 3.30 and 3.55, with no significant difference throughout the whole process. Meanwhile, the antioxidant activity of SJ was significantly enhanced by fermentation. For SW, the DPPH and OH· free radical clearance rates were significantly increased from 80.64% to 91.55% and from 44.72% to 97.10%, respectively. The reducing activity also had a significantly increasing trend from 3.79 to 3.94. He [19] reported the same trends during the production of SW. With the extension of fermentation time, the DPPH clearance rate and reducing activity improved significantly, reaching the maximum values at 6-8 d. Increases in levels of phenols and flavonoids were also found in the early stage of fermentation, which might be related to the improvement in antioxidant capacity. Wang et al. [25] found that at the concentration of 0.1 mg/mL, the DPPH radical scavenging activity of sea buckthorn tea leaves was 93.42%, indicating that sea buckthorn berry products have outstanding antioxidant activity. Nevertheless, distillation did not significantly decrease the antioxidant capacity ( Figure 2B). Wang and Li [26] also found that when the volume of aloe liquor was more than 2.5 mL, the scavenging rate of ·OH could reach more than 80%.

214
The comparison of VOCs of SJ, SW and DL is shown in Figure 3. PCA analysis 215 showed that the variance explained by the first principal component (PC1) and second 216 principal component (PC2) reached 75.7%, thus strongly reflecting the differences be-217 tween samples [27]. The three groups of samples were distinctly separated from each 218 other ( Figure 3A), indicating significant differences in VOCs among the three groups. A 219

Odour-Active Compounds
OAV is an index reflecting the contribution of one aroma to the overall flavour of samples. Table 1 lists the OAV values of VOCs in three groups of samples. There were 12, 16 and 17 odour-active components in SJ, SW and DL, respectively ( Figure 4A). An aroma component with OAV > 1 can be considered as a contributor to the overall flavour, and a component with OAV > 10 can be considered important [29].  [17]. Chauhan et al. [30] reported that the volatile aroma compounds of sea buckthorn berries were mainly aliphatic esters, such as ethyl esters of 3-methyl butyl and cis-3-hexen-1-yl. Esters usually have a fruity or floral odour, giving SJ a fresh and refreshing flavour [31,32].  Five VOCs increased in SW after fermentation, including benzeneacetic acid ethyl ester (OAV 2.610), geranyl acetone (6,10-dimethyl-5,9-undecadien-2-one, OAV 1.690), heptanal (OAV 1.500), heptanoic acid ethyl ester (OAV 1.182) and furfural (OAV 1.020). 4-Ethylphenol (OAV 25.400), phenylethyl alcohol (OAV 20.800) and nonanal (OAV 17.600) were added as important VOCs of SW. Phenylethanol with a rose fragrance is commonly found in wine, mangosteen and other fruit wines, and has been identified as a characteristic flavour compound. Ma [33] found that the relative content of phenylethanol in commercially sweet sea buckthorn wine was 7.193%. Lukša et al. [34] used Hanseniaspora uvarum yeast fermentation to produce sea buckthorn wine, and phenylethanol accounted for 3.6% of the overall aroma, indicating that phenylethanol is a stable flavour component in sea buckthorn wine. 4-Ethylphenol is usually present in rum and whiskey, and can be used as an essence in the preparation of wine and liquor [35,36]. Nonanal imparts a citrus fragrance (OAV 8.795) and contributes to the overall flavour of many wines [23].

Rapid Detection of Flavour
SJ, SW and DL samples were easily distinguished from each other by E-nose ( Figure 5A). Fermentation and distillation significantly increased the response values of SW and DL ( Figure 5B), indicating gradual enrichment of VOCs. The result is consistent with the findings of GC-MS. Correlation analysis was performed between the changes in E-nose values and each category of VOCs in the different groups ( Figure 5C). The results indicate that seven sensors, W5S, W6S, W1S, W1W, W2S, W2W and W3S, were positively correlated with increased alcohol and ester contents. SBJLp (lactic acid bacteria)-fermented SJ also represented high response values on W1W, W1S, W5S and W3S, similar to this experiment result [17]. W1C, W3C and W5C were negatively correlated with alcohol and ester contents, and positively correlated with acid contents. W3C and W5C reflected the increasing trends of carbonyl compounds, alkanes and terpenoids. In comparison with SJ, the response values of W3S, W2W, W2S, W1W and W1S for SW and DL were significantly increased, indicating that the alcohols and esters in the samples increased significantly after fermentation and distillation (especially the former). This result is consistent with the trends detected by GC-MS. The above findings show that E-nose could be used for rapid detection and evaluation of the changing trends of the main flavour compounds observed during the production of SW and DL. E-nose is not commonly used in flavour detection of sea buckthorn wine. Yu et al. [50] used E-nose to study the effects of different ultra-high-pressure treatments on the flavour and aging of SW, and found that treatment at 400 MPa significantly enhanced the aroma of SW. Some researchers have begun to establish correlations between GC-MS and E-nose data. For example, Long et al. [51] reported that S2, S6, S7 and S9 were important E-nose sensors for distinguishing between different cultivars of Alpinia officinarum based on the results of correlation analysis.

Correlation between E-Nose Values and Key Flavour Compounds
Correlation analysis was further conducted between the changes in response values of E-nose and 22 key flavour compounds ( Figure 5D). W5S, W6S, W1S, W1W, W2S, W2W and W3S values were also found to be positively correlated with 10 key flavour compounds,  OAV 132,666.7). W1C, W3C and W5C values were negatively correlated with the above 10 VOCs, indicating that these sensors reflected their downward trends. The above 10 VOCs all contributed to the overall flavour of sea buckthorn distilled liquor, particularly octanoic acid ethyl ester (OAV 16,466.7) and 1-(2,6,6trimethyl-1,3-cyclohexadien-1-yl)-2-buten-1-one (OAV 1132,666.7), indicating that the seven sensors (W5S, W6S, W1S, W1W, W2S, W2W and W3S) of E-nose can effectively reflect the strong aroma of sea buckthorn wine and distilled liquor in production. These results provide guidance for the rapid detection of the trends in key flavour compounds in SW and DL during production.

Fewer Differential Metabolites before and after Fermentation
Compared with sea buckthorn juice, seven significantly upregulated substances were found, which, although seemingly small in number, were important aroma substances; these substances were 3-octanone, phenylethyl alcohol, 2,3-butanediol, tridecane, octanoic acid ethyl ester, 6-methyl-5-hepten-2-one and acetoin. In addition, there were in fact 10 important flavour compounds that were newly produced after fermentation in SW, including 2-methyl-propanol, 3-ethoxy-1-propanol, 1-heptanol, benzyl alcohol and so on. All these VOCs contributed greatly to the overall flavour of sea buckthorn wine.
The OAV of phenylethanol in sea buckthorn juice was 10, and the OAV of phenylethanol in sea buckthorn wine increased to 20.8, indicating that phenylethanol occupied an increasing proportion in the overall flavour of sea buckthorn wine with the fermentation. Phenylethanol is formed by phenylalanine transamination of phenylpyruvate, followed by decarboxylation to phenylaldehyde, and phenylaldehyde in the presence of oxidative dehydrogenase generates β-phenylethanol [48]. Phenylethanol has a rose aroma, and is a favourite aroma in many fruit wines. Ma [34] and Lukša et al. [34] found that the relative content of phenylethanol detected in sea buckthorn wine was 7.193% and 3.6%, respectively, indicating that phenylethanol is a stable flavour component in sea buckthorn wine.
The OAV of octanoic acid ethyl ester in sea buckthorn juice was 1370, and the OAV of octanoic acid ethyl ester in sea buckthorn wine increased to 4110, indicating that octanoic acid ethyl ester was one of the key flavour compounds of sea buckthorn juice and wine, and its contribution was more obvious after fermentation. Ma [33] and Lukša et al. [34] also found that the relative content of octanoic acid ethyl ester detected in sea buckthorn wine was 0.179% and 0.12%, respectively, indicating that octanoic acid ethyl ester is also a stable flavour component in sea buckthorn wine. In this experiment, the content of octanoic acid ethyl ester in sea buckthorn wine was 0.411 ± 0.114 mg/L. Although the content seemed low, the threshold value of octanoic acid ethyl ester was low as 0.1 ug/kg; therefore, its OAV value was very high, and it had a very high contribution to the overall flavour of sea buckthorn wine.

Metabolic Pathways before and after Fermentation
Seven significantly upregulated and two downregulated VOCs were found in SJ after fermentation ( Figure 4B), and the metabolic pathways of fermentation were further analysed. In KEGG pathway analysis, the majority of metabolites were annotated as belonging to metabolic pathways and the biosynthesis of secondary metabolites (48 and 31 metabolites, respectively) ( Figure S1A). After fermentation, 15 upregulated pathways were found (log 2 FC > 1). The most obviously upregulated pathway was butanoate metabolism, followed by biosynthesis of cofactors, toluene degradation, furfural degradation, etc. ( Figure S1B). Fermentation significantly downregulated the fatty acid biosynthesis and lipoic acid metabolism pathways. Variable importance in projection (VIP) values were calculated for pathways, including the VOCs newly generated during fermentation (log 2 FC = Inf), of which five had VIP values > 1 ( Figure S1C): Butanoate metabolism, microbial metabolism in diverse environments, metabolic pathways, degradation of aromatic compounds and chemical-carcinogenesis DNA adducts.
Microbial metabolism in diverse environments and metabolic pathways are complex metabolic systems. Microbial metabolism in diverse environments includes carbohydrate metabolism, energy metabolism, metabolism of cofactors and vitamins and xenobiotic biodegradation. Metabolic pathways include carbohydrate metabolism, energy metabolism, lipid metabolism, nucleotide metabolism, amino acid metabolism, glycan metabolism, biosynthesis of terpenoids and polyketides, biosynthesis of other secondary metabolites and xenobiotic biodegradation.

Abundant Differential Metabolites before and after Distillation
Compared with the fermentation, the distillation process formed more flavour compounds. In total, 51 significantly upregulated and 14 downregulated VOCs were found in SW after distillation ( Figure 4C). In the present study, 51 growing flavour compounds were of interest, including 27 esters, eight carbonyl group compounds, four alcohols, three acids, two alkanes and seven other compounds. Especially, key flavour compounds such as linalool, isobutyl isovalerate, ethyl esters of octanoic acid and nonanoic acid showed obvious increasing trends. Therefore, many scholars have studied the changes of flavour compounds in the process of liquor distillation, and further take this as a benchmark to effectively control the distillation process in production.
Yang et al. [58] found that distillates aroma compounds have significantly differences in time periods in five-grain liquor. Concentration of aldehydes, esters and aromatics were increased in phase 1-2 (1∼10 min), decreased in phase 3 (10∼15 min), and reached the stable level after phase 3. Concentrations of alcohols decreased with the distillation time. Concentrations of 6-methyl-5-hepten-2-one, furfural, 3-furanmethanol and vanillin increased firstly, then decreased, and increased after phase 3. Liu et al. [59] also found that during distillation process of Fen-flavour liquor in China, the contents of acetaldehyde, acetal, isovaleraldehyde, sec-butyl alcohol, isobutanol, ethyl acetate, ethyl butyrate, isoamyl acetate, ethyl caproate and 2-pentanone decreased gradually with the progress of distillation. The contents of acetochlor, ethyl lactate, ethyl palmitate, acetic acid and isobutyric acid increased gradually with the progress of distillation. The contents of acetal, methanol and ethyl linoleate in initial distillate were significantly higher than other fractions. Further sensory evaluation results show that the sensory quality of first-round liquor was better than second-round, and the sensory quality of the early fraction was better than that of the later fraction. These results all provide data support for the "pinching the initial distillate and removing the last distillate" in actual production.
The association of GC-MS and E-nose needs to be an effective method to detect food flavour, considering that the detection accuracy of GC-MS is high, but it takes a long time, while E-nose detection is fast, but it cannot determine the specific substance. Huang et al. [60] integrated the volatile compounds and electronic nose response values of sugarcane juice under different treatments, and took Euclidean distance as the metric standard to generate the cluster analysis heat map of each index, and found that the response value of S6 may be related to 1-amyl alcohol and 1-octene-3-alcohol, and the response value of S8 and S10 may be related to the content of ethanethiol to a certain extent. Wu et al. [61] established the correlation between volatile compounds in the different dried Chrysanthemum nankingense and the response value of the E-nose sensor. Sensor W1C had a good correlation with C62 (hexyl n-valerate), sensor W5C had a good correlation with C136 (phenanthrene), and sensor W3C had a close correlation with C109 (geranialene). The sensor W1W had high sensitivity to alkenes, and had good correlation with various terpenes, such as C117 (caryophyllene) and C130 (1-isopropyl-4, 7-dimethyl-1,2,3,5,6,8 a-hexahydronaphthalene).

Conclusions
In this study, changes in the flavour of SW and DL at different processing stages were compared, the differential metabolites were analysed, and relationships between E-nose sensor values and key flavour compounds were established. A total of 133 VOCs were detected in the three groups, with 76, 79 and 99 VOCs identified in SJ, SW and DL, respectively. Sixteen and 17 VOCs contributed to the flavour of SW and DL, respectively. Fermentation significantly increased the contents of VOCs, especially esters. 3-Octanone and phenylethyl alcohol were the most significantly upregulated VOCs. E-nose data showed that seven sensors could reflect the increases in contents of alcohols, esters and 10 key flavour compounds.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/foods11203273/s1. Figure S1: Comparison of pathways in SJ, SW and DL by GC-MS (n = 3). (A) Main pathways by metabolites annotation, and log2FC (B) and VIP (C) for differential pathways after fermentation; Figure S2: Butanoate metabolism; Figure S3: Furfural degradation; Table S1: Concentration, p and VIP of VOCs of SJ, SW and DL.