Research on the application of liquid-liquid extraction-gas chromatography-mass spectrometry (LLE-GC-MS) and headspace-gas chromatography-ion mobility spectrometry (HS-GC-IMS) in distinguishing the Baiyunbian aged liquors

2.1 Samples of Baiyunbian liquor

Baiyunbian liquor is the typical representative of nong-jiang flavored liquors, which combines Maotai-aroma type liquor with Luzhou-aroma type liquor. Five Baiyunbian aged liquors samples (labelled as 3-year-storage liquors (3Y), 5Y, 12Y, 15Y and 20Y) were investigated by LLE-GC-MS and HS-GC-IMS, stored in ceramic wine jars and provided by Hubei Baiyunbian Co., Ltd. (HuBei province, China). All liquor samples (each in a volume of 500 mL and 42% of alcohol in each bottle) were stored at 4 °C until analysis.

2.2 Chemicals

Three types of extractants (freshly distilled before use), including methyl tert-butyl ether, n-pentane, and dichloromethane; analytical grade ethanol 99.7% (v/v), sodium hydroxide (NaOH), sodium chloride (NaCl), anhydrous sodium sulfate (Na₂SO₄) and sulfuric acid (H₂SO₄) were purchased from the Sinopharm Chemical Reagent Co., Ltd.

All chemicals are of analytical reagent grade, with at least 97% purity, purchased from Sigma-Aldrich Co., Ltd. (Shanghai, China). A C₈–C₄₀ n-alkane mixture (Sigma-Aldrich) was used for determination of retention indices (RIs).

3.1 Isolation of flavor compounds by LLE

According to the LLE method of Zheng with slight modifications [25], liquor sample (150 mL) was acidified to pH 1.0 with H₂SO₄ (6.0 mol/L) and then concentrated by a rotary evaporator of 50 °C to remove the ethanol (about 60 mL) in that the high concentration of ethanol will decrease the extraction effect of flavor compounds in samples, saturated with NaCl. The acidic liquor sample was extracted 3 times by freshly distilled extractant (20 mL each time). The combined extract recorded as No.1. Next, liquor sample was alkalined to pH 11.0 with NaOH (6.0 mol/L), saturated with NaCl and extracted 3 times by freshly distilled extractant. The combined extract recorded as No.2. In the third step, the liquor sample was neutralized to pH 7.0 with H₂SO₄ (1.0 mol/L), saturated with NaCl, and extracted 3 times by freshly distilled extractant. The combined extract recorded as No.3. Finally, the three samples were dried by Na₂SO₄ overnight, concentrated to a final volume of 0.5 mL under a gentle stream of nitrogen, and then stored at −20 °C until analyzed by GC-MS.

3.2 Preparation of simulated liquor samples

The simulated liquor samples (42%(v/v) ethanol), containing a serious of known concentrations of standard samples: Ethyl butyrate, ethyl valerate, ethyl hexanoate, ethyl octanoate, propanol, 2-methylpropanol, butanol, pentanol, propionic acid, butyric acid, hexanoic acid, octanoic acid, furfural and tetramethylpyrazine were operated in accordance with the steps (3.1) of the real liquor samples for method validation. In order to simulate the real liquor sample as much as possible, the pH values of the simulated liquor samples were adjusted to 3.7 with H₂SO₄.

3.3 GC-MS analysis

Identification of aroma compounds was performed on an Agilent 7890B GC, coupled with an Agilent 5977B MS (Agilent Technologies, USA). The DB-Wax column (30 m × 0.25 mm ID, 0.25 m film thickness) was used to perform the chromatographic separations. Helium was used as the carrier gas with a constant flow mode of 0.7 mL/min. The injector temperature was kept at 260 °C. The oven temperature program was hold at 45 °C for 1.5 min, raised to 85 °C at a rate of 6 °C, then, sequentially increased at a rate 4 °C/min to 225 °C for 15 min, along with quadrupole mass filter was at 150 °C, the transfer line temperature was at 250 °C and ion source temperature at 230 °C. The MS parameters included electron impact ionization with electron energy of 70 eV, and mass range of m/z 50–550, using the selective ion monitoring (SIM) mode. The area of each peak was determined by MassHunter software (Agilent Technologies).

3.4 Identification and quantification of flavor compounds

Identification of flavor compounds was carried out by comparison with retention index (RI), and NIST 11 library (Agilent Technologies). RI was calculated using the C₈–C₄₀ n-alkane mixture under the same GC-MS condition described above. The two methods were used for comprehensive qualitative analysis of compounds.

Methyl tert-butyl ether was used as the organic solvent, amyl acetate (internal standard sample) and each standard sample were dissolved into the solvent to form a mixed standard sample, and then injected on GC-MS. Then take the peak area as the ordinate and the concentration as the abscissa to quantify the compounds.

The information of all chemicals including Compound Name, CAS number, Retention time (Rt), Quantifier ion (m/z), Qualifier ion (m/z) was contained in Table 1.

3.5 GC-IMS analysis

Measurements were made on a GC-IMS (FlavourSpec®, Gesellschaft für Analytische Sensorsysteme mbH, Dortmund, Germany). The GC was equipped with a WAX capillary column (30 m × 0.53 mm). 1 mL of sample was poured into 20 mL headspace vials and heated at 60 °C for 10 min. Then, 100 µL aliquot of the headspace was automatically injected into the heated injector (65 °C) in splitless mode. The GC column operated at a constant temperature (60 °C) and N₂ (99.99%) was used as sample gas at a flow rate of 150 mL/min and as a drift gas at a programmed flow as follows: 2 mL/min for 20 min, 80 mL/min for 10 min. The drift gas entered the device in the opposite direction to the ions, in order to prevent non-ionised molecules from entering the ionisation chamber. The analytes were eluted and separated at 45 °C, and then ionized in the IMS ionization chamber by a ³H ionization source in a positive ion mode. The 9.8 cm drift tube was driven at 45 °C in a positive drift voltage mode. Finally, the data was recorded during the analysis for 30 min.

3.6 Statistical analysis

Statistical procedures were carried out using the SPSS version 13.0 statistical package for windows (SPSS Inc., Chicago, Ill., U.S.A.). The data of five samples from LLE-GC-MS and HS-GC-IMS were analyzed by cluster analysis and principal component analysis (PCA), respectively, to investigate the possible differences among the five aged liquors.

4.1 Optimization of LLE method

Five parameters including types of extractants, pH values in acidic and alkaline environment, volume of extractant, and rotary evaporation temperature were investigated in this work, based on the average peak areas and the number of compounds.

According to the reports [26], the number of compounds and recovery rate are different due to different organic extractants. The extraction efficiency of dichloromethane, chloroform, n-pentane, n-hexane and ethyl acetate were been extensively studied [27]. The results showed that the extraction efficiency of ethyl acetate and alkanes was lower (20–30%). Dichloromethane and chloroform can obtain better extraction effects (70–90%), and the polarity of dichloromethane is closer to the flavor substances in liquors. According to previous studies that we have done in preliminary experiments, methyl tert-butyl ether has a good extraction effect. Therefore, the extraction efficiency of dichloromethane, n-pentane and methyl tert-butyl ether were compared in this study, aim at extracting more flavor compounds.

According to Figure 1B, 48, 40, and 62 compounds of aged liquor were identified by dichloromethane, n-pentane and methyl tert-butyl ether, respectively. In Table 2, it was concluded that there exists large differences among various compounds extracted by the three extractants. Except for aromatics and other compounds, the P values of the other seven classes of compounds were all less than 0.05. Especially for the three classes of main compounds in the sample: esters, alcohols and acids, the P values of them were all less than 0.01, indicating the three classes of main compounds had the largest differences among the three extractants. As can be seen from Figure 1A, there were 6 esters, 6 alcohols and 1 acid extracted by methyl tert-butyl ether, but didn’t be extracted by dichloromethane; there were 7 esters, 11 alcohols, and 3 acids extracted by methyl tert-butyl ether, but didn’t be extracted by n-pentane. The average peak areas of same compounds (Figure 1C) obtained by methyl tert-butyl ether were higher than the other two extractants. As described above, methyl tert-butyl ether was selected as the optimal extractant to further study. However, there are no reports about methyl tert-butyl ether in extracting flavor substances of liquor until now. Therefore, methyl tert-butyl ether is the innovation point of this research.

Figure 1:

The number of various compounds (A) and total compounds (B), and peak area response values (C) extracted by three extractants.

The pH value optimization in acidic and alkaline environment, the extractant volume, and the rotary evaporation temperature optimization were similar to the extractants optimization described above. The optimized five parameters of LLE were as follows: methyl tert-butyl ether as extractant, acidic environment of C_H⁺ = 1 mol/L, alkaline environment of pH = 12, extractant of 30 mL and rotary evaporation temperature of 50 °C.

4.2 Method validation

The validation of the method was performed by measuring the detective linearity, limits of detection (LODs), limits of quantification (LOQs), repeatability, reproducibility and recovery of compounds.

To set up a standard curve, the stock solution was diluted into five different concentrations levels. Three replicate working standard solutions at each concentration level were detected, in order to obtain an average peak area of compounds that was set as the ordinate, while the known concentration level of compounds was set as the abscissa. As shown in Table 3, the coefficient of variation (CV, %) of the five different concentrations, in the range of 1.4–9.6, was considered to be great method accuracy. The linear correlation coefficients were great (R² > 0.99) except for 2-methylpropanol (0.9863) and octanoic acid (0.9816). LODs and LOQs were used to assess the sensitivity of the method based on a signal to noise ratio of 3:1 and 10:1, respectively. The ranges of LODs and LOQs were 0.11–0.51 μg/L and 0.33–1.78 μg/L, respectively, well below the concentration of all compounds in the samples, which implied that the method was accurate and results were reliable. The repeatability of the method was detected by the relative standard deviation (RSD%) of a series of six replicates performed on the same day. The RSD varied between 5.21 and 14.67% with an average value of 10.54%. The reproducibility of the method [28] was tested by different operators on the same equipment at different days, repeated six times. RSD fluctuated from 7.67 to 17.71% with an average value of 11.15%. They were all operated at stable temperatures at laboratory, in order to ensure a reliable repeatability and reproducibility. As shown in Table 4, the simulated liquor samples were extracted by the optimized method to obtain the relative recovery, performed in triplicate. The each compound showed very good recoveries (77.8–113.4%) with standard deviation ranged from 3.8 and 18.4%, which was in ideal range for flavor analysis.

4.3 Application of the optimized LLE method for different aged liquors

There are thousands of compounds in baijiu. According to the structure of the compounds, the main flavor components in Liquor can be divided into esters, alcohols, acids, aldehydes, ketones, phenols, nitrogen-containing compounds, sulfur-containing compounds, etc. [29]. Owing to their different concentration and proportion of flavor compounds [30], baijiu is quite different in quality and taste.

Take the chromatogram of 3Y sample as representative, as shown in Figure 2. A total of 119 flavor compounds (Figure 3) were identified in five samples, including 51 esters (42.86%), 16 alcohols (13.45%), 12 acids (10.08%), 6 aromatics (5.04%), 11 nitrogen-containing compounds (9.24%), 1 sulfur-containing compound (0.84%), 7 furans (5.88%), 5 ketones (4.2%), and 10 other compounds (8.4%). Among them (Figure 4A), the number of esters increased with the storage years, but did not increase obviously in 15Y. The alcohols did not show a stable change. There were 10 acids in 3Y, 5Y, and 12Y; and 11 acids in 15Y and 20Y. Only one sulfur-containing compound was detected in five kinds of aged liquors: dimethyl disulfide. Both nitrogen-containing compounds and furans increased with the growth of the storage years. As shown in Figure 4B, the number of total flavor compounds increased with the cellaring ages. Esters, alcohols and acids were the most abundant in liquors, followed by nitrogen-containing compounds and furans.

Figure 2:

The chromatogram of 3Y on gas chromatography-mass spectrometry (GC-MS).

Figure 3:

The proportion of various flavor compounds.

Figure 4:

The change tendency of various compounds (A and D) and total compounds (B and C) in different storage period.

As shown in Figure 4C, the total flavor concentration of 20Y was the highest (4543.23 mg/L), and the 3Y was the lowest (3984.96 mg/L).Therefore, the total flavor concentration increased with the growth of the cellaring ages. The concentrations of most compounds are greatly different, which may be an important factor in forming the unique quality of various aged liquors. The result is consistent with Wang’s conclusion [31]. In Table 5, it was concluded that there exists large difference among various compounds of five aged liquors. Except for aromatics and other compounds, the P values of the other seven classes of compounds were all less than 0.05. Among them, the P values of esters, alcohols, acids and nitrogen-containing compounds were all less than 0.01, indicating the four classes of compounds had the significant differences among the five aged liquors.

As shown in Figure 4D, the largest group was esters in liquor samples, which was consistent with the results of Xiao et al. [32]. They mainly contribute to flowery, fruity and sweet aroma compounds to the Liquor [33]. The total ester concentrations (Figure 4D) were 2041.18 mg/L (3Y), 2132.43 mg/L (5Y), 2343.1 mg/L (12Y), 2534.67 mg/L (15Y) and 2580.38 mg/L (20Y). The esters increased with the cellaring ages, indicating that acids and alcohols had been undergoing esterification, and the speed of esterification reaction was faster than the hydrolysis reaction. But they did not increase apparently in 20Y sample that have been stored for more than 15 years, which showed that the esterification reaction get slow and the aged liquor gradually matured after 15 years [34]. Among them, ethyl hexanoate had the highest concentration (1354.16–1660.52 mg/L) in all esters, followed by ethyl lactate (577.98–659.04 mg/L) and ethyl butyrate (33.21–151.1 mg/L) whose concentration increased with the cellaring ages. The concentration of the remaining esters was less than 30 mg/L, of which methyl formate > ethyl valerate > ethyl octanoate > ethyl heptanoate > 10 mg/L. The highest concentration of the same compound in the five aged liquors was recorded as 5, and the lowest was recorded as 1. The relative ratio relationship of the seven esters in five different storage years was shown in Figure 5A. The concentration of methyl formate was the highest in the 3Y, followed by 5Y. The concentration of Ethyl octanoate was the highest in the 12Y, followed by 15Y. In the 20Y, the concentration of ethyl hexanoate, ethyl 2-hydroxypropionate, ethyl butyrate, ethyl heptanoate and ethyl valerate was the highest, followed by 15Y.

Figure 5:

Rader plots of esters (A), alcohols (B), acids (C) and pyrazines (D) in five aged liquors.

In this study, the total alcohol concentrations (Figure 4D) were 1337.26 mg/L (3Y), 1347.33 mg/L (5Y), 1222.21 mg/L (12Y), 1247.93 mg/L (15Y), and 1278.79 mg/L (20Y), which did not keep an absolutely stable change but showed a general downward trend. The main reasons for the reduction of alcohols were they can be oxidized to aldehydes, esterified to esters and condensed to acetals with the cellaring ages [34], [35]. Among them, the concentration of propanol was highest (the highest concentration was 934.73 mg/L in 3Y), followed by 3-methylbutanol (the highest concentration was 222.13 mg/L in 20Y). Butanol > 100 mg/L > 2-methylpropanol > pentanol > hexanol > 10 mg/L, the remaining alcohols was less than 10 mg/L. The highest concentration of the same compound in the five aged liquors was recorded as 5, and the lowest was recorded as 1. The relative ratio relationship of the six alcohols in five different storage years was shown in Figure 5B. In the 3Y, the concentration of propanol and 2-methylpropanol were the highest, followed by 5Y. In the 15Y, the concentration of pentanol was the highest, followed by 20Y. In the 20Y, the concentration of hexanol, 3-methylbutanol and butanol were the highest. Alcohols could mainly be produced during the fermentation under aerobic conditions from sugars and amino acids. Small amounts of alcohols could also be produced by yeast through the reduction of the corresponding aldehydes [36]. Among them, hexanol contributes to floral and green aromas, and isoamyl alcohol gives a nail polish aroma. Pentanol and butanol contribute to fruity odour and an alcoholic aroma, respectively [33].

The total acid concentrations (Figure 4D) were 470.87 mg/L (3Y), 494.54 mg/L (5Y), 542.08 mg/L (12Y), 520.16 mg/L (15Y), and 497.73 mg/L (20Y), which increased from 3Y to 12Y mainly caused by the aldehydes being oxidized to acids, and decreased from 12Y to 20Y proving that the esterification reaction was faster than the oxidation reaction from aldehydes to acids [34]. Among them, the concentration of hexanoic acid was highest (the highest concentration was 458.34 mg/L in 12Y) in acids, followed by propionic acid (the highest concentration was 44.25 mg/L in 20Y) and butyric acid (the highest concentration was 19.09 mg/L in 12Y). Octanoic acid > valeric acid > 10 mg/L, the remaining acids was less than 10 mg/L. The highest concentration of the same compound in the five aged liquors was recorded as 5, and the lowest was recorded as 1. The relative ratio relationship of the five acids in five different storage years was shown in Figure 5C. In 5Y, the concentration of hexanoic acid was second only to the 12Y. In the 12Y, the concentration of valeric acid, hexanoic acid and butyric acid were highest. In the 15Y, the concentration of octanoic acid was highest, followed by 20Y. In the 20Y, the concentration of propionic acid was highest. Acids were produced during bacterial spoilage but were always formed by yeasts during fermentation [37]. Acids are responsible for fruity, cheese, fatty and rancid notes. Short-chain acids which have the aroma of sour, rancid could suppress and cover the other aroma in Liquor and have strong adhesion that the taste organs feel a long-lasting stimulating effect during the drinking. So the appropriate concentration of acids is essential to the balance and harmony of the overall flavor [38].

In the study, nitrogen-containing compounds were mainly pyrazines. Pyrazines are characteristic flavor ingredients of nong-jiang flavored liquor, which was the fourth largest compound in the liquor. The total pyrazine concentrations (Figure 4D) were 71.35 mg/L (3Y), 76.48 mg/L (5Y), 80.78 mg/L (12Y), 94.62 mg/L (15Y), and 99.43 mg/L (20Y), which increased with the cellaring ages owing to Maillard reactions between sugars and amino residues [39] and they rose rapidly in 12–15 years, which was the turning point of the change of total concentration of pyrazines. The study was consistent with the results of jing’s research [40]. There were 4 pyrazines quantified in five aged liquors, including 2-ethyl-6-methylpyrazine, 2,3,5-trimethylpyrazine, tetramethylpyrazine, and 2-ethyl-3,5-dimethylpyrazine. Among them, the tetramethylpyrazine (TMP) was most abundant (35.32–43.31 mg/L) in all pyrazines, which can reduce the incidence of cardiovascular and cerebrovascular disease, inhibits platelet aggregation, and protects liver [41]. Then, 2,3,5-trimethylpyrazine > 2-ethyl-6-methylpyrazine > 10 mg/L > 2-ethyl-3,5-dimethylpyrazine. The highest concentration of the same compound in the five aged liquors was recorded as 5, and the lowest was recorded as 1. The relative ratio relationship of four pyrazines in five different storage years was shown in Figure 5D. The concentrations of four pyrazines were the highest in 20Y, followed by 15Y. The results showed that the concentration of pyrazines was directly related to cellaring years of liquor samples [42]. “The liquor needs to be stored for a long time before it becomes very fragrant” was clearly reflected in the pyrazines.

Only one sulfur-containing compound was detected in five kinds of aged liquors: dimethyl disulfide, which mainly contributes to aroma characteristics of cabbage and onion [43]. Its concentrations in 3Y, 5Y, 12Y, 15Y, and 20Y were 11.34, 12.21, 10.8, 9.48, and 10.21 mg/L, respectively, that showed a downward trend with cellaring ages, due to the unstable molecular structure of volatile sulfur-containing compounds [44]. The total furan concentrations (Figure 4C) were 50.91 mg/L (3Y), 52.31 mg/L (5Y), 59.89 mg/L (12Y), 68.59 mg/L (15Y), and 72.96 mg/L (20Y), which increased with the cellaring ages. The accumulation of furans is generally considered a sign of aging, as it can be formed by pyrolysis of carbohydrates, Maillard dehydration of sugars, and caramelization during the fermentation and storage.

According to many studies, ceramic wine jars contain numerous metal ions and small voids that distribute on the inner wall of jars [45]. Alcohols can be oxidized into aldehydes after fully contacting the oxygen molecules in small voids and the reaction speed can be accelerated with the participation of metal ions. Then the aldehydes are further oxidized to acids. Then, acids can promote the formation of esters [34]. Meanwhile, esters also produce alcohols and acids through hydrolysis reaction. The dynamic balance of acids and esters can be gradually formed with the increase of cellaring ages [34]. During storage, aldehydes and alcohols can also form acetals by condensation reaction [35]. In addition, trace sugars and amino residues can gradually produce nitrogen-containing compounds through Maillard reaction [39].

Five aged liquors were visualized and naturally grouped by cluster analysis. The squared Euclidean distance and Ward clustering technique was showed in dendrogram. The result of cluster analysis to the flavor compounds was shown in Figure 6. The similarity of the sample, revealed by LLE-GC-MS, was highest between 15Y and 20Y, followed by 3Y and 5Y. The cluster analysis results of the five aged liquors coincided with the storage time of the liquor samples.

Figure 6:

Dendrogram of cluster analysis in five aged liquors.

4.4 The characteristic flavor fingerprints of five samples by HS-GC-IMS

For further understanding the flavor profiles in aged liquors, HS-SPME-GC-MS was used to analyze flavor compounds. The emphasis of HS-GC-IMS analysis was focused on the fingerprints-based, non-targeted analytical method for discriminating differences of five aged liquors. As previous researches, the information of the whole spectral fingerprint was taken into account to compare the differences in aromas compounds [46]. The drift time was normalized relative to the reaction ion peak (RIP) position. 80% of data was focused on the area of the topographic plot (Figures 7A, B): retention time extended from 180 to 800 s and drift time ranged from 1.0 to 1.9 s.

Figure 7:

Topographic plots (A and B), gallery plot (C) of five aged liquors.

For further studies, 60 signals (45 known compounds and 15 unknown compounds (Figure 7C) were used to set up the gallery plot through comparing the differences of topographic plots. The apparent differences among the aged liquors were clearly noticed. A single compound was corresponding one or more signals because of adducts formed between the ions and neutral molecules (e.g., dimers, trimers) during moving through the drift tub, which demonstrate detected results were relate to proton affinity and concentration of compounds [47], [48]. As shown in the red area, ethyl 2-methylbutyrate, butyl hexanoate, ethyl isobutyrate, 1-hexanol, 2-pentanone, furfural, valeraldehyde, 2,3-butanedione, compound 11(C11), C12, C13, C14, and C15 increased with the cellaring ages. 2-methylpropyl acetate, isobutyl acetate, propyl acetate, ethyl acetate, ethyl formate, acetic acid, 2-methyl-1-propanol, 2-octanol (blue area) decreased with the storage years.

PCA plots try to simply explain the maximum amount of data by decreasing the dimensionality to two-dimensional space without much loss of information [49]. It was used to confirm the main sources of variables, in order to reveal the classification among the samples [50]. In general, the maximum variable information in all original data can be interpreted by the two first principal components, in terms of PC1 and PC2, respectively [28]. PCA of 60 flavor compounds among different aged liquors was shown in Figure 8. The accumulative variance contribution rate of the first PC (57.0%) and the second PC (14.0%) accounted for 71.0%. However, many previous works [51], [52], [53] also pointed out that the cumulative percentage of the first principal component PC1 and the second principal component PC2 in the PCA diagrams is above 70% still can analyze the experiment results.

Figure 8:

Principal component analysis of five aged liquors.

Aged liquors samples of 3Y (−2400 PC1 and −1200 PC2), 5Y (−1800 PC1 and 900 PC2), 12Y (600 PC1 and 1000 PC2), and 20Y (1900 PC1 and −800 PC2) were settled down into the third, second, first and fourth quadrant, respectively. 15Y (1900 PC1 and 0 PC2) was located in between the first and fourth quadrant that overlaps with 20Y in some areas. The differences of flavor compounds were shown by different scores of PC1 and PC2 in five samples. The results revealed that 3Y, 5Y, and 12Y were remarkably distinguished according to the flavor compounds. However, the flavors compounds of the 15Y and 20Y samples were very similar and cannot be completely distinguished. Therefore, the selected compounds could be as the biological markers used for differentiating different aged liquors.

In short, the characteristic aromas fingerprints of aged liquors were successfully established through HS-GC-IMS, which clearly observed distinguish of five aged liquors. The differences among the five samples were obvious, which was similar with the results of LLE-GC-MS.