Evaluation of Perceptual Interactions between Ester Aroma Components in Langjiu by GC-MS, GC-O, Sensory Analysis, and Vector Model

The volatile compounds of three Langjiu (“Honghualangshi, HHL”, “Zhenpinlang, ZPL”, and “Langpailangjiu, LPLJ”) were studied by gas chromatography-olfactometry (GC-O) and gas chromatography-mass spectrometry (GC-MS). The results showed that a total of 31, 30, and 30 ester compounds making a contribution to aroma were present in the HHL, ZPL, and LPLJ samples, respectively. From these esters, 16 compounds were identified as important odour substances, and their odour activity values (OAVs) were greater than 1. The key ester components were selected as: ethyl acetate, ethyl 2-methylbutyrate, ethyl 3-methyl butyrate, ethyl hexanoate, and ethyl phenylacetate by aroma extract dilution analysis (AEDA), odour activity value (OAV), and omission testing. Five esters were studied for perceptual interactions while using Feller’s additive model, OAV, and a vector model. Among these mixtures, they all have an enhancing or synergistic effect. Among these mixtures, one mixture presented an additive effect and nine mixtures showed a synergistic effect.


Introduction
Chinese liquor, which is a popular alcoholic beverage in China, is composed of ethanol, water, and trace components. Ethanol and water account for 98% to 99% of the total mass of liquor, and the remaining 1% to 2% is composed of trace compounds. It is this 1% to 2% (the trace components) that can determine the flavour and style of the liquor. Chinese liquor consists of three aroma types according to the diversity of aromas: a sauce-aroma, strong-aroma, and light-aroma. Langjiu is sauce-aroma type of liquor. They are made from sorghum, wheat, etc., and they are fermented, distilled, aged, and blended by traditional solid-state methods [1]. The content of esters in different flavoured liquors varies, and the general ester components account for 35% to 70% of the total aroma components [2].
In recent years, research into the aroma of liquor has been reported: Cheng et al. [3] used a headspace-solid phase microextraction (HS-SPME)-mass spectrometry (MS) technique, partial least squares discriminant analysis (PLS-DA), and stepwise linear discriminant analysis (SLDA) methods to determine 131 Chinese liquor samples. Finally, 32 characteristic ions were selected, and 32 ions were then input; eight groups of ions of different geographical origins were used as outputs to establish an artificial neural network (ANN) recognition model. Liu et al. [4] introduced the direct analysis of Langjiu and its serial products by WH-3 glass chromatographic column. The application of such methods could rapidly detect more than 20 kinds of main flavour compositions, including acids, esters, Bedford, MA, USA). Sodium chloride (analytical grade) and absolute ethanol (analytical grade) were obtained from Sino-pharm Chemical Reagent Co., Ltd (Shanghai, China).

Extraction of Volatile Compounds of Langjiu by Headspace Solid-phase Microextraction (HS-SPME)
The volatile compounds were extracted by HS-SPME, as follows: three liquor samples were diluted with deionised water to a 10% ethanol concentration. We added 8 mL liquor sample, 1.5 g sodium chloride, and 50 µL internal standard (2-octanol, 400 mg/L) to the 15 mL headspace bottle that had a PTFE−silicone septum, and then put the headspace bottle in a constant temperature water bath at 50 • C. A 50/30 µm divinylbenzene/carboxyl/polydimethylsiloxane (DVB/CAR/PDMS) fiber was exposed in the headspace without stirring for 50 min., and then desorbed into the injection port of the gas chromatograph for 5 min. At the end of each analysis, the fiber was inserted into a thermal heater at 250 • C for 20 min. to ensure that there was no residue. Each liquor sample went through the same process, as described above.

Identification by GC-MS and GC-O
GC-MS analysis was conducted on a 7890 gas chromatograph (GC) coupled to a 5973C mass (MS) (Agilent Technologies, Santa Clara, CA, USA). GC-O analysis used an Agilent 7890A gas chromatograph (GC), which was equipped with a Gerstel ODP2 detector (Mülheim a der Ruhr, Germany).

GC-MS Analysis
The liquor samples were analysed while using two types of columns: an HP-Innowax column (60 m × 0.25 mm × 0.2 µm; Agilent) and a DB-5 column (60 m × 0.25 mm × 0.25 µm; Agilent). Using helium (purity 99.999%) as a carrier gas, the flow rate was 1 mL/min. The quadrupole mass filter has a temperature of 150 • C and a transfer line temperature of 250 • C [9]. The oven temperature was set to 40 • C (6 min), ramped at 3 • C/min. to 100 • C, and then ramped at 5 • C/min. to 230 • C (20 min). Mass spectrometry conditions were as follows: electron ionisation (EI) mode at 70 eV (ion source temperature 230 • C) was used and the scan range was m/z 30-450. Volatile components were identified by comparing the retention index (RI), molecular weights, and mass fragmentation patterns in the database (Wiley 7n.L Database, NIST Database) to authentic standards.

GC-O Analysis
After the liquor sample enters the gas chromatograph, it was separated by the chromatographic column and then flowed to the detector and olfactory orifice at 1:1, respectively. The chromatographic columns were an HP-Innowax (60 m × 0.25 mm × 0.25 µm; Agilent) and a DB-5 (60 m × 0.25 mm × 0.25 µm; Agilent). Using hydrogen as the carrier gas, the flow rate was 2mL/min. The oven temperature was set to 40 • C (6 min), ramped at 3 • C/min. to 100 • C, and ramped at 5 • C/min. to 230 • C (20 min). The injector temperature was set to 250 • C and the FID temperature was set to 280 • C. In addition, the moist air entered the sniffing port at a flow rate of 50 mL/min. to expel residual aroma compounds from the sniffing port [10]. Each aroma compound was determined by comparing the RI, the odour, and mass spectra of the standard products. The FD coefficient represents the maximum dilution coefficient of each compound ( Table 1). All of the trials were carried out on each liquor sample three times.

Aroma Extract Dilution Analysis
For AEDA, the liquor samples were diluted with deionised water, and the sample with an ethanol content of 10 (v/v) was obtained. The sample was gradually diluted with 10% ethanol and water (1:1) until reaching 1:1024. Each dilution was submitted to GC-O analysis under the same GC conditions that are described above until no odorant was detected. The flavour dilution (FD) factor of each compound represented the maximum dilution at which the odorant could be perceived. The identification of each aroma compound was conducted by comparing their odours, RI, and mass spectra with those of pure standards. All of the trials were carried out on each liquor sample three times.

Quantitative Analysis
Thirty-one aroma compounds were quantified from the calibration curves. Using the prepared model liquor sample, the standard substance of appropriate concentration was added, and then diluted into six concentration gradients in turn, each concentration gradient point was extracted and then analysed three times, followed by the addition of internal standard solution (2-octane, 50 µL, 400 mg/L) to establish the calibration curves of the aroma substance for determining the aroma. It was used to determine the actual concentration of aroma substances in each liquor sample. The extraction conditions for solid phase microextraction (SPME) were the same as those of Langjiu. Table 2 lists the coefficients of the calibration curves, where y represents the peak area ratio (peak area of volatile standard/peak area of internal standard) and x denotes the concentration ratio (concentration of volatile standard/concentration of internal standard).

General Conditions
Sensory analysis was performed on behalf of Martin and Revel [11] (1999). The 10 mL sample was placed in a brown glass bottle, randomly numbered while using three digits, and then evaluated in different compartments at room temperature (20 • C).

Sensory Panels
The assessors were grouped into sensory panel A (10 males and 10 females) and sensory panel B (two males and two females). Sensory panel A participated in the determination of threshold and model establishment, and sensory panel B participated in the determination of the dilution factor by GC-O. Sensory panels consisted of 24 volunteers (12 males and 12 females, aged between 23 and 29 years). The volunteers were selected from 40 candidates based on their performance in several olfactory tests. They suffered no problems, such as olfactory allergies. All of the volunteers were from the School of Perfume and Aroma Technology, Shanghai Institute of Technology. They attended meetings twice a week for four weeks.

Omission Analysis
Triangular tests were performed for selecting the key esters of Langjiu. The panellists attended meetings twice a week for 1.5 hours each for three weeks. Triangular omission tests for key esters in Langjiu: only one compound was omitted (Table 4; tests 1 to 14) from the 14 esters, and then compared with the samples of all the key esters (14 esters). The ester concentration was the actual concentration of the ester in Honghualang (with an ethanol level of 53% (v/v)). In the triangulation test, each group had to randomly arrange three coded samples: one different sample and two identical samples. All of the liquor samples were shown to volunteers three times. The volunteer selected samples containing aroma compounds in three samples, although they were unsure. The results were based on published, tabulated data and were statistically analysed according to the binomial law of the distribution of answers in such tests.

Determination of Odour Thresholds and OAVs
Through the omission experiment, the selected key esters were mixed in pairs, and the olfactory threshold of the binary mixture was measured in an aqueous solution of 53% ethanol and was conducted while using three alternative forced selection tests (3-AFC). The OAV was used to determine the contribution of aroma substances to the aroma of the liquor. The OAV was the ratio of the concentration of aroma substance to the threshold of the substance.

Determination of Intensity of Binary Mixtures
Water solutions of 1-butanol were prepared at 25 ± 1 • C, according to the odor intensity referencing scale (OIRS, from level 1 (aqueous solution of 10 ppm) to level 12 (20,480 ppm)). The binary mixtures of ethyl acetate and ethyl hexanoate, and ethyl acetate and ethyl 2-methylbutyrate were mixed at the same strength, and the strength of the mixture was determined. The experiment was repeated three times.

Feller's Additive Model
The olfactory threshold of mixed aroma substances was established. The results of all three alternative forced selection tests were statistically analysed. The results were summarised and presented as a detection probability and detection confidence of chemical stimulus. The detection probability was given by: where P = detection probability corrected for chance, m = number of choices per trial (this article, three), and p(c) = proportion correct (number of correct trials/total number of trials). The sigmoid (logistic) equation was employed to model the psychometric function for groups and each individual, as follows: where c is olfactory threshold of the odorant (log µg/L), where x represents odorant concentration (log µg/L), where P is detection probability (0 ≤ P ≤ 1), and D is a parameter characteristic of each odorant that defines the gradient of the function [12][13][14].
Feller's additive model could be used to evaluate the interactive effects of the mixtures [14]. The actual model that was obtained from the mixture experiment was compared to a simple additive theoretical model. The detection probability formula of the mixture P(AB) was as follows: where P(A) represents the probability of detecting component A and P(B) is the probability of detecting component B. If the sum of probabilities was higher than the panel's detection performance for the mixture, some degree of suppression had occurred relative to statistical independence, in accordance with statistical independence, a certain degree of inhibition had occurred. On the contrary, some form of mutual addition or synergy had occurred. Furthermore, no mixing interaction occurred if the sum of probabilities matches was equal to the detection performance.

The Odour Activity Value Approach
Ferreira V. [15] proposed that the odour activity values (OAVs) or concentration/threshold ratios of the odorant mixture at the threshold between the binary mixtures were approximately additive. That is, if a mixture contains n odorants and the sum of n concentrations/thresholds is y, then the mixture is above the threshold by y times. In arithmetic form: wherein the OAVmix refers to the number of times that the mixture was diluted to reach the threshold, and OAVi was the proportional concentration/threshold of the ith odorant of the mixture (the threshold was measured separately). OAVmix was originally defined as the threshold of the mixture and was recorded as Tm. Subsequently, as Ti and Ci were the thresholds and concentrations of the ith component of the mixture, respectively, individual OAVi values were calculated, added, and divided by the threshold of the mixture. This parameter was called X: X values of 1 represent odour additivity, while a reduced value represents increased interaction or synergy. X values greater than 1 means that antagonism occurs [16].

The Vector Model
The vector model could be thought of as an adjacent edge of a parallelogram, where the length of the edge represents the perceived intensity of the unmixed component, while the length of the diagonal in the figure represents the perceived intensity of the mixture [17]. Therefore, the OI of the binary mixture was successfully correlated with the odour intensity of its unmixed components, as follows: where a and b represent two different substances, and OI ab is the OI of a mixture of a and b. The interaction coefficient cosα (where α is the angle between the sides of the parallelogram) represents the degree of interaction between the two unmixed components of the binary odour mixture.
In general, different odour mixtures had different values of cosα, which were usually based on experience to determine the components with equal perceptual intensity and they were used to predict the OI of the remaining mixtures in a group. For special cases where the perceptual intensities of the two odour components were equal, Equation (6) can be rewritten, i.e., OI a = OI b , and the following equation applies [18]: The vector value (OIab) can be used to replace the actual aroma intensity of the mixture since the vector model is a perfect predictor of the aroma intensity of the mixture.

Statistical Analysis
Analysis of variance (ANOVA) analysed the concentration of volatile compounds, and the interaction of esters in the Feller's additive model and the vector model was analysed by Sigma Plot 12.0 (SYSTAT) software (Systat Software Inc, San Jose, CA, USA). The level of statistical significance was 5% (p < 0.05).

Qualitative and Quantitative Analysis
The qualitative and quantitative analysis of esters in langjiu was carried out to more accurately reveal the perceptual interaction between esters in Langjiu. Through GC-O sniffing and identification analysis, 31 ester compounds were found in the three kinds of Langjiu, application of GC-O to the liquors revealed 17, 17, and 16 aroma compounds (FD ≥ 16) in HHL, ZPL, and LPLJ, respectively ( Table 1). The differences in the number of aroma compounds (FD ≥ 16) were mainly caused by concentration differences. These aroma compounds were determined by comparison with authentic standards, retention indices, and aroma descriptions. HHL contains more aroma substances, among which ethyl hexanoate (1024) [19]. These esters were mainly formed by the metabolism of yeast, filamentous fungi, etc., or fatty acid esterification reaction during fermentation [20]. Table 2 shows the concentrations and relative deviations of these compounds in Langjiu. Among these esters, ethyl acetate (450,892-529,294 µg/L) was the most abundant, followed by ethyl lactate (340,025-428,330 µg/L); in addition, ethyl propionate (32,654-35,598 µg/L), ethyl butyrate (23,585-27,387 µg/L), ethyl hexanoate (6078-13,849 µg/L), and ethyl 3-methyl butyrate (11,795 µg/L) were also present in higher concentrations. Wei, L. and Zhang, L. [5] used dichloromethane as the extractant for determining the main aroma components of Langjiu by gas chromatography-mass spectrometry (GC-MS). A total of 31 trace components were identified, and the most abundant were: ethyl hexanoate, hexanoate acid, ethyl lactate, acetic acid, and butyric acid. This was slightly different from the research results of Wei and others. This might have been due to the different extraction methods used to isolate aroma substances.

Threshold and OAV of Ester Compounds in Langjiu
Although GC-O analysis was an effective means of aroma compound identification, it did not accurately indicate the contribution of aroma compounds to the overall aroma. In liquor samples, aroma substances at a concentration above the detection threshold also contribute to the overall aroma. Therefore, individual OAVs were calculated to assess the contribution of different aromatic compounds to the aroma [21].
The aroma activity values of 24 ester aroma compounds in Langjiu were calculated by referring to the smell threshold of aroma substances in alcohol solution in the literature, and based on the quantitative results in different kinds of Langjiu. Table 3 shows the OAV calculation showed that the aroma contribution of each compound. It was found that 16 [8], and these esters were also key aroma substances in maotai-flavour liquor [22].  (Table 4; tests 1 to 14) among 14 esters, so that a sample containing all of the studies compounds (14 esters) was compared with that only omitting one compound. For each group in triangulation tests, three coded samples were randomly arranged: one different and two identical. Through the omission testing of each compound, the results showed that these compounds had a significant effect on the overall aroma of the ester mixture. For ethyl acetate, ethyl 2-methylbutyrate, ethyl 3-methyl butyrate, ethyl hexanoate, and ethyl phenylacetate, the results showed that the difference was significant with p < 0.001. This was inconsistent with the conclusions of Fan et al. [25], because the liquor used and the pre-treatment methods were inconsistent. Table 4. Olfactory impact of the omission of various esters from complex aromatic reconstitutions. C 2 C 2 C 3 C 2 C i4 C 2 C 4 C 2 MeC 4 C 2 C i5 C 2 C 2 C i5 C 5 C 2 C 6 C 2 C 8 C 2 C 7 C 2 C 2 BeC 2 PrBeC 2 2OHC 3 C 2 Difference Observed

Selection of Five Ester Aroma Compounds
The results showed that esters made a significant contribution to the overall aroma of liquor [22]. Furthermore, through the study of the ester compounds in three kinds of Langjiu, GC-MS and GC-O technology identified 31 ester compounds. The key aroma components were further screened by omission test (p < 0.001). Five key esters were selected, respectively, ethyl acetate (p < 0.001), ethyl 2-methylbutyrate (p < 0.001), ethyl 3-methyl butyrate (p < 0.001), ethyl hexanoate (p < 0.001), and ethyl phenylacetate (p < 0.001). Finally, the perceptual interaction between the five esters was studied by using Feller's additive model, odour activity values, and a vector model.

Olfactory Properties of Compounds
It is unreasonable to consider the overall aroma of Langjiu as the sum of the flavour contributions of each compound, because the interaction of different senses affecting flavour perception will be ignored, although the threshold of aroma compounds can be used as an indicator of their influence on flavour. Therefore, the interaction between substances was studied through the change of threshold before and after mixing.

Studying the Olfactory Properties of Compounds by Feller's Additive Model
The change of threshold between the binary mixtures of key esters was revealed, and the experimental results were analysed, to investigate the interaction between the binary mixtures. The interaction between aromatics was studied by Feller's additive model.
The five key ester compounds screened by omission experiment were mixed according to the proportion of their actual concentration in HHL. A total of ten groups of compounds were used to study the interaction of key ester compounds, namely: acetate and ethyl 2-methylbutyrate mixed, ethyl acetate and ethyl 3-methyl butyrate mixed, ethyl acetate and ethyl hexanoate mixed, ethyl acetate and ethyl phenylacetate mixed, ethyl 2-methylbutyrate and ethyl 3-methyl butyrate mixed, ethyl 2-methylbutyrate and ethyl hexanoate mixed, ethyl 2-methylbutyrate and ethyl phenylacetate mixed, ethyl 3-methyl butyrate and ethyl hexanoate mixed, ethyl 3-methyl butyrate and ethyl phenylacetate mixed, and ethyl hexanoate and ethyl phenylacetate mixed. The probability of detection of the binary mixture could be calculated by Feller's additive model, and then estimated by the Feller model threshold, as derived from the sigmoid (logistic) equation.
The detection probabilities calculated using Feller's addition model were lower than the actual detection probabilities that were obtained by the experiment (Figure 1 The results showed that the mixture approaches the response-addition model at low detection levels, i.e., the independence of the assay, while they approach the dose-addition model at a high detection level. The ratio of the actual detection threshold obtained by the experiment and the theoretical threshold calculated by the Feller additive model, the lowest ratio of ethyl 2-methylbutyrate and ethyl phenylacetate was 0.10 (Figure 1g), and the highest ratio of ethyl acetate and ethyl 2-methylbutyrate was 0.57, according to the experimental results (Figure 1a). It could be seen from Figure 1 that the interaction between different aromatic compounds was different, which might be due to various factors such as the molecular size of the aromatic compounds themselves, the types of functional groups and their own volatility, as well as different intermolecular van der Waals forces and hydrogen bonds [28].  The ratio of the actual detection threshold obtained by the experiment and the theoretical threshold calculated by the Feller additive model, the lowest ratio of ethyl 2-methylbutyrate and ethyl phenylacetate was 0.10 (Figure 1g), and the highest ratio of ethyl acetate and ethyl 2-methylbutyrate was 0.57, according to the experimental results (Figure 1a). It could be seen from Figure 1 that the interaction between different aromatic compounds was different, which might be due to various factors such as the molecular size of the aromatic compounds themselves, the types of functional groups and their own volatility, as well as different intermolecular van der Waals forces and hydrogen bonds [28].

Studying the Olfactory Properties of Compounds by the OAV Approach
The OAV has been applied to a large number of binary, ternary, and more complex mixtures, so it could be used to confirm the interaction between key ester compounds in Langjiu. Binary mixture OAVmix and ∑OAVi were calculated while using Equation (4), and the difference between the two was compared using Equation (5). Table 5 shows the experimental results. The ten groups of mixtures all have X < 1, the ethyl acetate and ethyl 2-methylbutyrate mixed was an additive effect, and the other groups shows a synergistic, which was consistent with results from Feller's additive model.

Studying the Olfactory Properties of Compounds by the OAV Approach
The OAV has been applied to a large number of binary, ternary, and more complex mixtures, so it could be used to confirm the interaction between key ester compounds in Langjiu. Binary mixture OAVmix and OAVi were calculated while using Equation (4), and the difference between the two was compared using Equation (5). Table 5 shows the experimental results. The ten groups of mixtures all have X < 1, the ethyl acetate and ethyl 2-methylbutyrate mixed was an additive effect, and the other groups shows a synergistic, which was consistent with results from Feller's additive model. Many researchers have conducted extensive research into the OAV approach: Guadagni et al. [29,30] studied compounds containing nitrogen and sulphur in potatoes and found that these compounds have different effects on the overall aroma of the potato. Laura et al. [31] studied nine important oxidation-related aldehydes while using the OAV approach, revealing the interaction (addition or synergy) with other volatile compounds in wine. For example, the ratio of OAVmix to OAVi of a mixture of (E)-2-hexenal, (E)-2-octenal, and (E)-2-nonenal was 3, which showed a synergistic effect. In addition, the (E)-2-enoyls were negatively correlated with the flavour of red wine, while branched aldehydes could enhance the dryness of the fruit and mask the negative effects of (E)-2-alkenals on the flavour of red wine.

A Vector Model of Perceptual Odour Interaction
A binary mixture of ethyl acetate and ethyl hexanoate and a binary mixture of ethyl acetate and ethyl 2-methylbutanoate were selected by vector model since the vector model can distinguish the interaction between the two mixtures. Yan et al. [17] used a vector model to study the relationship between binary mixtures of aldehydes and ester binary mixtures, and the results evinced good correlation. Ethyl acetate is similar in structure to ethyl hexanoate. Five groups of equal interaction ethyl acetate and ethyl hexanoate were used for binary mixing and the strength of the mixture was determined. cos 1 2 α was calculated according to Equation (7), and the binary substances were then obtained. The interactive relationship ( Figure 2) is such that cos 1 2 α = 0.8072. The ethyl acetate and ethyl 2-methylbutanoate with different structures were selected for analysis. The result revealed that cos 1 2 α = 0.6577. The vector model can directly study the interaction between aroma substances, which was helpful in finding the law of interaction between aroma substances.

Conclusions
Qualitative and quantitative analyses of volatile esters in three kinds of Langjiu by GC-O and GC-MS with headspace solid phase microextraction (HS-SPME) were undertaken. A total of 31 ester compounds were identified, and 31 of them were quantitative analysis. FD value (FD ≥ 16), OAV (OAV ≥ 1), and omission test screened the key esters, and the results showed that ethyl acetate, ethyl 2-methylbutyrate, ethyl 3-methyl butyrate, ethyl hexanoate, and ethyl phenylacetate contributed to the aroma of Langjiu to a significant extent. Through the study of the interaction of binary mixtures in key esters by Feller's additive model, OAV, and a vector model, it was confirmed that these ester compounds had additive or synergistic effects. Trace aroma components in liquor, especially the esters, have great influence on the flavour and quality of liquor, which is one of the important bases

Conclusions
Qualitative and quantitative analyses of volatile esters in three kinds of Langjiu by GC-O and GC-MS with headspace solid phase microextraction (HS-SPME) were undertaken. A total of 31 ester compounds were identified, and 31 of them were quantitative analysis. FD value (FD ≥ 16), OAV (OAV ≥ 1), and omission test screened the key esters, and the results showed that ethyl acetate, ethyl 2-methylbutyrate, ethyl 3-methyl butyrate, ethyl hexanoate, and ethyl phenylacetate contributed to the aroma of Langjiu to a significant extent. Through the study of the interaction of binary mixtures in key esters by Feller's additive model, OAV, and a vector model, it was confirmed that these ester compounds had additive or synergistic effects. Trace aroma components in liquor, especially the esters, have great influence on the flavour and quality of liquor, which is one of the important bases to judge the quality of liquor. The experimental results provide a scientific basis for the analysis and determination of liquor flavour substances and the evaluation of sensory quality, and they are of guiding significance for the improvement of liquor fermentation technology to improve the aroma quality of liquor.