Received: 10 March 2015, Revised: 30 November 2015, Accepted: 1 December 2015 Published online in Wiley Online Library: 7 January 2016

Transcription

1 Research article Received: 10 March 2015, Revised: 30 November 2015, Accepted: 1 December 2015 Published online in Wiley Online Library: 7 January 2016 (wileyonlinelibrary.com) DOI /ffj.3304 Characterization of different aroma-types of chinese liquors based on their aroma profile by gas chromatography mass spectrometry and sensory evaluation Zuobing Xiao, a,b Dan Yu, a,b Yunwei Niu, a *NingMa a and Jiancai Zhu a Abstract: The aroma profile of different aroma-types of Chinese liquors was measured using gas chromatography mass spectrometry (GC-MS) and sensory evaluation. Sensory analysis coupled with principal component analysis (PCA) showed that strong-aroma liquors were mainly characterized by cellar aroma; light-aroma liquors were mainly characterized by fen, floral, grain, sweet and vinegar aromas while sauce-aroma liquors were mainly characterized by sauce, fruity and caramel aromas. Partial least squares regression (PLSR) results indicated that ethyl hexanoate, hexanoic acid, hexyl hexanoate, 4-methyl pheno, ethyl pentanoate, isoamyl hexanoate and pentanoic acid were significantly and positively associated with cellar aroma and strongaroma liquors. Vinegar and fen aroma were strongly linked with light-aroma liquor and associated with of ethyl acetate, 2-methylpropyl acetate, isoamyl acetate, isoamyl lactate, ethyl decanoate, diethyl butanedioate, 2-phenethyl acetate, 3,4-dihydro-2H-1-benzopyran-2-one 1-dodecanol, ethyl decanoate, 3-methyl-1-butanol and thymol. While sauce-aroma liquors were positively correlated with caramel, fruity and sauce sensory descriptors and volatile compounds such as ethyl 2-methylpropanoate, ethyl 3-methylbutanoate, furfural, 1,3-dimethyl trisulfide, trimethyl pyrazine, ethyl benzeneacetate, benzaldehyde and 1-octanol. Keywords: Chinese liquor; SBSE/GC-MS; volatile compounds; PLSR Introduction Chinese liquor is one of the leading distillates in the world. According to the statistics from China National Association for Liquor and Spirits Circulation, an accumulative consumption of 12 million kiloliters and sales revenue of about 5000 billion RMB are obtained in 2013 in China. Chinese liquors are composed of three aroma type liquors, including sauce-aroma, strong-aroma and light-aroma types, based on the diversity of aroma. Variation in aroma and taste among different aroma types can be attributed to the liquor production processes of fermentation, duqu (usually used as inoculum for the solid-state fermentation), [1,2] distillation, [3] blending and aging. It is essential to characterize the aroma composition and sensory properties of Chinese liquor and to quantitatively assess their variability among aroma types to ensure the consistent quality control of liquor products. Gas chromatography mass spectrometry (GC-MS) is a comprehensive technique for identification and quantitation of volatile and semi-volatile composition of aroma compounds in liquor. Wenlai Fan et al. [4 6] have investigated the aroma compounds of strong-aroma liquors (yangheduqu, Wuliangye and Jiannanchun) by GC-MS and GC-Olfactometry. It showed the most important aroma compounds in Chinese liquors could be ethyl butanoate, ethyl pentanoate, ethyl hexanoate, ethyl octanoate, butyl hexanoate, ethyl 3-methylbutanoate, hexanoic acid, and 1,1-diethoxy-3-methylbutane. Shukui Zhu et al. [7] developed a method to characterize the volatile compounds in Moutai liquor (sauce-aroma liquor) by comprehensive two-dimensional gas chromatography/time of-flight mass spectrometry and 528 components were identified. Wenlai Fan et al. [8] characterized the volatile compounds in Chinese soy sauce aroma type liquor by stir bar sorptive extraction and gas chromatography mass spectrometry (SBSE/ GC-MS). However, no comparison of the volatiles and aroma sensory profiles of the three aroma type Chinese liquors has been performed. The aroma profile, which endows the distinctive aroma features, is positively correlated to the classification of Chinese liquors. So it is necessary to explore the potential correlation between sensory attributes and aroma compounds by multivariate statistical analyses, such as principal component analysis (PCA), and partial least squares regression (PLSR). It can also be valuable techniques to characterize liquors among different aroma types. However, few reports are available about sensory evaluation and aroma composition among different aroma type liquors. The objective of this study were to (1) investigate the aroma and volatiles of different aroma type liquors by sensory evaluation and GC-MS; (2) evaluate the relationship between aroma compounds, sensory attributes and different aroma type liquors. Through the above research, the comparison of the volatiles and aroma sensory profiles of several aroma type Chinese liquors were elucidated. We expect this study will help to set up the foundation for the * Correspondence to: Yunwei Niu, Haiquan Road 100, Fengxian, Shanghai, China yunweiniu@163.com a School of Perfume and Aroma Technology, Shanghai Institute of Technology, Shanghai , PR China b Shanghai Research Institute of Fragrance and Flavor Industry, Shanghai , PR China 217

2 Z. Xiao et al. 218 characterization of Chinese liquors and the discrimination among different aroma types. A better understanding of this knowledge will be helpful for the effective improvement of characteristic aroma of different aroma-types of Chinese liquors through adjusting fermentation and aging parameters. Experimental Liquors and materials Three different aroma types of Chinese liquor manufactured by Chinese liquor companies were evaluated in this study. 16 Chinese commercial liquors were selected, among which seven were representative of strongaroma type, five of light-aroma type and four of sauce-aroma type. Seven strong-aroma liquors were purchased commercially from Wuliangye Yibin Co., Ltd. (St1, Wuliangye, 500 ml, 52% ethanol by volume), Anhui Gujing Group Co., Ltd. (St2, Gujing, 500 ml, 42% ethanol by volume), Sichun Langjiu Group Co., Ltd. (St3, Langpeitequ, 500 ml, 42% ethanol by volume), Jiangsu Yanghe Distillery Co., Ltd. (St4, Yanghe, 500 ml, 55% ethanol by volume), Beijing Shunxin Agriculture Niulanshan Distillery Co., Ltd. (St5, Niulanshan, 500 ml, 50% ethanol by volume), Shandong Qufu Confucius Family Liquor Co.,Ltd.(St6,Kongfujia,500mL,52%ethanolbyvolume),andShanxiXifeng Liquor Co., Ltd. (St7, Xifeng, 500 ml, 50% ethanol by volume), respectively. Five light-aroma liquors were purchased from Beijing Red Star Co., Ltd. (Li1, Hongxing, 500 ml, 53% ethanol by volume), Henan Baofeng Wine Co., Ltd. (Li2, Baofeng, 500 ml, 52% ethanol by volume), Kinmen Kaoliang Liquor (Xiamen) Trading Co., Ltd. (Li3, Kaoliang, 500 ml, 58% ethanol by volume) and Shanxi Xinghua Fen Group Co., Ltd. (Li4, Xinghuacui, 500 ml, 53% ethanol by volume; Li5, Fenjiu, 500 ml, 53% ethanol by volume), respectively. Four sauce-aroma liquors were purchased from Guizhou Sauce Collar Wines Ltd. (Sa1, Jiangling, 500 ml, 53% ethanol by volume), Sichun Langjiu Group Co., Ltd. (Sa2, Laolangjiu, 500 ml, 53% ethanol by volume), Hunan Wuling Spirits Co., Ltd. (Sa3, Wulingjiu, 500 ml, 53% ethanol by volume), and Kweichow Moutai Co., Ltd. (Sa4, Miaotai, 500 ml, 53% ethanol by volume), respectively. In this study all samples were diluted with pure water to a lower level of ethanol matrix (10%). All experiments were performed in triplicate. Polydimethylsiloxane (PDMS)-coated stir bars (Twister) of 10 mm length and 0.5 mm film thickness and glass vials (20 ml) were obtained from Gerstel, Mullheima/d Ruhr, Germany. Chemicals Authentic standards were obtained from Sigma-Aldrich (St. Louis, MO, USA): ethyl acetate, ethyl propanoate, propyl acetate, 2-methylpropyl acetate, ethyl 2-methylbutanoate, ethyl nonanoate, ethyl trans-4-decenoate, 2-pentanol, 2-methyl-1-butanol, 2-heptanol, 2,6-dimethyl-4-heptanol, 1-octen-3-ol, 2-nonanol, benzyl alcohol, 2-phenethyl alcohol, acetic acid, butanoic acid, hexanoic acid, octanoic acid, dodecanoic acid, 1,1-diethoxy ethane, 2-methylbutanal, hexanal, heptanal, octanal, nananal, decanal, 2-heptanone, 3-octanone, 6-methyl-5-hepten-2-one, 2-nonanone, 2-decanone, 2-undecanone, phenol, thymol, trimethyl pyrazine, tetramethyl pyrazine, furfural, 5-methyl-2- furfural and dimethyl disulfide. Authentic standards were obtained from Fluka (Buchs, Switzerland): ethyl butanoate, ethyl 3-methylbutanoate, isoamyl acetate, ethyl pentanoate, pentyl acetate, butyl butanoate, ethyl hexanoate, hexyl acetate, ethyl heptanoate, ethyl lactate, ethyl octanoate, hexyl hexanoate, ethyl decanoate, diethyl butanedioate, ethyl benzeneacetate, 2-phenethyl acetate, 2-butanol, 1-butanol, 3-methyl-1-butanol, 1-pentanol,1-hexanol, 3-octanol, 1-heptanol, 2-ethyl-1-hexanol, 1-octanol, 1-nonanol, 1-dodecanol, heptanoic acid, decanoic acid, benzaldehyde, (2E)-2-nonenal, (E,E)-2,4- decadienal, 2-octanone, 2-ethyl phenol and 1,3-dimethyl trisulfide. Authentic standards were obtained from TCI America (Portland, OR, USA): ethyl 2-methylpropanoate, isoamyl butyrate, propyl hexanoate, 2-undecanol, 3-methyl butanoic acid, pentanoic acid,1,1-diethoxy hexane, 1,1,3,3- tetraethoxypropane, 1,1-diethoxy nonane, 2-methoxy-4-methyl phenol, 4-ethyl-2-methoxy phenol, 4-methyl phenol and 2-n-butyl furan. 2-octanol was used as internal standards, purchased from Sigma-Aldrich (St. Louis, MO, USA). Purity of chemical standards is over 95% in all cases, most of them are over 99%. Pure water was obtained from a Milli-Q purification system (Millipore, USA). Sodium chloride (analytical grade) was purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). Sensory evaluation The aroma profile of Chinese liquors was evaluated by a well-trained panel consisting in 10 members (24 40 old), four females and six males, all of them belonging to the laboratory staff and with a long experience in sensory descriptive analysis. The sensory analysis method, which has been previously described [9,10] was conducted according to the ISO 8589 standard procedure and the China National Institute of Standardization (CNIS) GB/T with minor modifications. Five sensory sessions, each spent 1.5 h, were conducted. In the three preliminary sessions, the panelists thoroughly discussed the aroma properties of liquors and nine attributes (sauce, cellar, fen, fruity, floral, sweet, grain, caramel and vinegar) according to the objective of the present work were generated and described for further descriptive analysis. In sessions four and five, panelists were asked to rate the liquor samples in random orders on a 10-point intensity scale (0 = absence, 1 = very low, and 9 = very high ) for each attribute. A constant volume of 15 ml liquor was evaluated in 50 ml volume wine-taster glasses at room temperature, in individual booths, coded with three-digit random numbers, under clean air conditions. A gap of 20 s was used to separate two individual odour assessments [11] and clean air was breathed between each assessment. Each liquor sample was rated three times for each attribute. SBSE for extraction of volatile compounds The extraction of volatile compounds by SBSE has been optimized and the optimized procedure was as follows: to a 20 ml glass vial was added 10 ml of diluted liquor sample (10% ethanol content), 10μL of an internal standard solution (2-octanol, 400 mg L 1 in ethanol) and 2 g of sodium chloride. A preconditioned stir bar was introduced to the vial, which was stirred at 750 rpm in a 4 agitation point plate (Meiyingpu Instrument Inc., Shanghai, China). The equilibrium time has been elaborated in a previous paper. [12] Then the extraction temperature was set at room temperature (25 C, thermostatted room)withextractiontimeof90min. Following extraction, the coated stir bar was removed from the matrix and rinsed with distilled water to remove ethanol and salt. Then the coated stir bar was gently wiped with a clean lint-free tissue to remove water, followed by loading it into a blank desorption tube (180 mm length, 4 mm OD, 3 mm ID, Gerstel). The analytes were thermally desorbed from the stir bar by introducing the desorption tube into the Gerstel Thermo Desorption System (TDS3) and subjected to GC-MS analysis. The recondition was performed by introducing the thermal desorption tube to the Gerstel tube conditioner TC at 300 C under a constant flow of nitrogen (75mL min 1 )for1h. GC-MS analysis of volatiles and calculation of odour activity values A 7890 GC coupled to a 5973 MS (Agilent Technologies, Palo Alto, CA, USA) equipped with a thermal desorption system (TDS3) and a programmed temperature vaporizing injector (CIS4), both from Gerstel, were used throughout the study. For thermal desorption, the TDS was operated in the solvent vent mode for the first 3 min and splitless mode thereafter. The transfer capillary temperature was kept constant at 280 C. Desorption temperature was programmed from 40 (equilibrium for 1 min) to 260 C (held at this temperature for 5 min) at 60 C min 1 under a helium flow (75 ml min 1 ) and the desorbed analytes were cryofocused in the PTV system with liquid nitrogen at 90 C. Finally, the PTV system was programmed from 90 (equilibrium for 1 min) to 250 C (held for 1 min) at 12 C s 1. wileyonlinelibrary.com/journal/ffj

3 Characterization of chinese liquors by GC MS and sensory evaluation Separation was performed on an Agilent 7890 GC-5973 MS system, equipped with a HP innowax capillary column (60 m length, 0.25 mm ID and 0.25μm film thickness, Agilent Technologies). The carrier gas was helium at a flow rate of 1.0 ml min 1. The GC oven programme was as follows: 40 C held for 2 min, ramped at 4 C min 1 to 230 C (held for 5 min). For mass spectrometry analysis, electron impact mode (EI) at 70 ev was used and the MS transfer line and ion source temperature were 280 and 230 C, respectively, with the MS scanning from 30 to 400amu. Identification of the volatile compounds was achieved by comparing mass spectra and retention indices (RI) with reference compounds when they were available. Tentative identification was performed based on matching Mass Spectral Library (NIST08, Wiley7n. l) and comparison of RI with those previously reported in literatures when standard compound was not available. For calculation of RI a C7-C30 n-alkanes series (concentration of 1000 mg L 1 in n-hexane) from Sigma-Aldrich was used. To obtain a matrix similar to that of Chinese liquor, model wine was prepared containing standard compounds in 10% of ethanol of Milli-Qdeionized water. The ph was adjusted to 3.2. Quantification of the major volatile compounds was conducted by calibration curves obtained by each volatile compound from the model solution. Five levels of concentration for the calibration were performed in triplicate. 50μL 2-octanol (400 mg L 1 in ethanol, internal standard) was spiked into the 10 ml of model wine in a 20mL vial and then extracted by SBSE. Then the average peak area ratios (peak area of a compound to the internal standard) were used as y axis whereas x were the known concentration of standard. For the compounds with no available standard, the concentration of the volatile compounds in Chinese liquors was determined by using the calibration curves of a similar functional group standards, the respective standards were shown in Table 2. Statistical analysis Analysis of variance (ANOVA) was performed to find significant differences among volatile compounds of Chinese liquors on different aroma types. For the data obtained in sensory descriptive analysis, ANOVA was performed with the attribute intensity scores derived from a panel of judges. Duncan s multiple range tests were conducted when the samples exhibited significance between them, with the level of significance set at P < Both ANOVA and Duncan s multiple range tests were conducted by the SAS V8 (SAS Institute Inc., Cary, NC, USA). Principal component analysis (PCA) was run using XLSTAT ver (Addinsoft, New York, NY, USA). PCA was conducted to extract important information from sensory analysis data to evaluate the aroma profile similarity between three aroma type liquors, and also to identify the specific factors leading to the greatest variability. To determine the potential correlation between volatile compounds with (X-matrix) and aroma sensory attributes (Y-matrix) among liquor samples, partial least squares regression (PLSR) [13] was conducted with the Unscrambler version 9.7 (CAMO ASA, Oslo, Norway). All variables were centred and standardized (1/SD) in order to obtain normalized data, and an unbiased contribution of each variable to the criterion would be obtained as each variable had a unit variance and zero mean before applying PLS analyses. Full cross-validation was carried out to validate PLSR models. To determine the optimum number of PLSR components included in the analysis, the smallest prediction error sum of squares was used. Results and discussion Sensory evaluation of three aroma-type chinese liquors The sensory attributes and the definition for each attribute were provided in Table 1. Examination of the results of sensory analysis based on ANOVA and Duncan s multiple range tests indicated that all sensory attributes, with the exception of fruity, were found to have significant differences (p < 0.05) among the three types of liquor. The results were then plotted in a spider web diagram, as shown in Figure 1a. Strong-aroma liquors had the highest scores in cellar aroma, but had lowest scores in sweet and vinegar aromas. Light-aroma liquors received the highest scores in fen, floral, grain, sweet and vinegar aromas and had a high level of cellar aroma. Sauce-aroma liquors were scored highest in sauce, fruity and caramel aromas and lowest in cellar and fen aromas. These characters separated the three type liquors from each other. It suggested that different aroma type liquors varied greatly in their aroma profiles. Principal component analysis (PCA) was used to identify the relationships between the Chinese liquors studied and sensory attributes. Figure 1b showed the bi-plot for the first two principal components, which accounted for 62.51% of the total variance. The loading of PC1 contributed for the most parts of cellar, grain, sweet and vinegar aromas whereas PC2 were positively characterized by caramel and sauce, and negatively associated with fruity, floral and fen aromas. The strong-aroma liquors (coded with St1-St7) were all located at the left side, which were primarily characterized by cellar aroma. Light-aroma liquors (coded with Li1-Li5) were all located at positive values of PC1 and negative values of PC2. This type of liquor was positively correlated with grain, fruity, floral and fen aromas. Sauce-aroma liquors were located at positive values of PC1 and PC2, characterized by the contribution of caramel and sauce aromas. Volatile composition of liquors In the present study, a total of 111 volatile compounds were detected using SBSE/GC MS in Chinese liquors, among which 89 volatile compounds were identified based on the comparison of their MS and RI with reference compounds, whereas 22 volatile compounds were identified by matching with MS library and comparison of their RI with those reported in literatures. 63 compounds were found in all aroma type liquors. Table 2 presented the average and standard deviation for each compounds in Chinese liquors. 66 volatile compounds showed significant differences (P < 0.05) among aroma types as a consequence of differences in winemaking processes. Esters were the most prevalent class in terms of the number of volatiles in Chinese liquors, with ethyl esters of the linear fatty acids being presented at the highest concentration, particularly ethyl acetate, ethyl butyrate, ethyl hexanoate, ethyl pentanoate, ethyl heptanoate, ethyl octanoate. These esters were mainly the markers of fermentative aroma [14] and contributed to the pleasant fruity aroma of liquors. These esters, except for ethyl octanoate, all exhibited significantly different levels (p < 0.05) among the liquors. Ethyl hexanoate had been reported as the most abundant compound in strong-aroma liquors. As expected, it showed significant highest concentrations in strong-aroma liquors ( mg L 1 ) in this study, while in sauce-aroma and light-aroma liquors it varied a lot, the concentration was mg L 1,6.33mgL 1,respectively. Light-aroma and sauce-aroma liquors possessed the higher level of ethyl acetate, mg L 1, mg L 1, respectively, followed by strong-aroma liquors with 5.39 mg L 1.Itisidentical with our common understanding that ethyl acetate is the main component of light-aroma liquors. [15] The acetate esters, including isoamyl acetate, hexyl acetate, 2-phenylethyl acetate, also conferred the liquor aroma their fruity nuances. It was found that their quantified amount in Chinese liquor was similar with that of white wines. [16] Ethyl butanoate, ethyl pentanoate and ethyl heptanoate showed the significantly higher concentration in both strongaroma and sauce aroma liquors, but less in light-aroma liquors. The observed differences among the three aroma types can be 219 wileyonlinelibrary.com/journal/ffj

4 Z. Xiao et al. Table 1. The sensory attributes, description and reference standard used in the study Attributes Description Reference Standards Caramel The aroma of baking cereals and grains and caramel 5g crushed caramel in 100 ml 10% ethanol water solution Cellar The aroma characteristics of ethyl hexanoate 1mg L 1 10% ethanol water solution of ethyl hexanoate Fen The aroma characteristics of ethyl acetate 1mg L 1 10% ethanol water solution ofethyl acetate Fruity Exhilarating scent and smelling strongly of varieties of ripe fruits 1mg L 1 10% ethanol water solution ofethyl acetate, ethyl butanoate, ethyl hexanoate and ethyl octanoate Floral The perfume of fresh flowers 1mg L 1 10% ethanol water solution of 2-phenyl ethanol Grain The steamed food aroma from the - fermentation and distillation of Sorghum, rice, wheat and other grains Sauce The aroma characteristics of traditional soybean paste sauce products Sweet Pleasant candy-like and honey-like note Sweet candy Vinegar The aroma characteristics of acids Shanxi Mature Vinegar 220 ascribed to variation in grain composition during harvest, resulting from differences in climatic conditions and alcoholic fermentation. 21 higher alcohols were identified and quantified. Among them, 10 alcohols varied significantly among aroma types (p < 0.05). 3- methyl-1-butanol and 2-phenethyl alcohol appeared with the highest concentrations in all liquors. These two alcohols were previously found to represent more than 80% of the volatile fraction of Garnacha Tintorera based sweet wine [17] and Godello white wines. [18] 3-methyl-1-butanol reached the highest concentration (7.48 mg L 1 ) in light-aroma liquors. Noguerol-Pato et al., [19] found that 3-methyl-1-butanol was quantitatively (249 mg L-1) the most abundant higher alcohol in Mencía wines, reminiscent of alcohol, fusel odour. 2-phenethyl alcohol reached the highest concentration (14.01 mg L 1 ) in sauce-aroma liquors. 10 different fatty acids were identified in the present study. Among them, significant differences were observed except for nonanoic acid, decanoic acid and dodecanoic acid. In small amounts, fatty acid can contribute to a balanced aroma in liquors by hindering hydrolysis of their esters. [20] Hexanoic acid exhibited the highest concentration of mg L 1, above 20 mg L -1[21,22] in strong-aroma liquors. It may be negative and confer undesirable odour to liquors. A total of eight acetals were quantified in this study. 1,1-diethoxy ethane, 1,1-diethoxy-3-methylbutane, 1-(1-ethoxyethoxy)-pentane and (2,2-diethoxyethyl)-benzene showed significant differences among liquors. (2, 2-diethoxyethyl)-benzene, the most abundant acetals, reached the highest concentration in both light-aroma (22.31 mg L 1 ) and sauce-aroma liquors (22.21 mg L 1 ). Carbonyl compounds were present in liquors at relatively low concentrations. Among aldehydes, 2-methylbutanal displayed the highest concentration in sauce-aroma liquors. Other aldehydes appeared only with trace amounts, far lower than 2-methylbutanal. This probably ascribed to the fact that they can be reduced to the corresponding alcohols in the process of fermentation. [16] Strong-aroma liquors had the highest concentration of 2-nonanone (1.16 mg L 1 ). Light-aroma liquors had the highest concentration of 3,4-dihydro-2H-1-benzopyran- 2-one (1.37mgL 1 ) and dihydro-5-propyl-2(3h)-furanone (1.79mg L 1 ). The group of volatile phenols was relatively large, which included six compunds, 2-methoxy-4-methyl phenol, phenol, 4-ethyl-2-methoxy phenol, 4-methyl phenol, 4-ethyl phenol and thymol. Only 4-methyl phenol and thymol showed significant differences among liquors. Light-aroma liquors possessed the highest amounts of volatile phenols, especially 2-methoxy-4-methyl phenol (7.30mg L 1 ) and 4-ethyl-2-methoxy phenol (6.07mg L 1 ). Another group was the pyrazine compounds, with a total of four pyrazines quantified in the present study. These pyrazines displayed highest concentrations in sauce-aroma liquors, which could be related to the Maillard reaction between saccharide and amino residues or the ambient temperature reaction of microbial metabolites in the solid-state fermentation in Chinese liquor. [23] Five furans were identified, 2-n-butyl furan, furfural, 5-methyl-2-furfural, 3,4-dihydro-2H-1- benzopyran-2-one and dihydro-5-propyl-2(3h)-furanone. 2-n-butyl furan and furfural showed significant differences among liquors and reached the major values in sauce-aroma liquors. 5-methyl-2-furfural was only detected in sauce-aroma liquors with higher concentration of 1.74 mg L 1. This was agreed with the previously reports for Chinese soy sauce aroma type liquor (sauce-aroma liquor) by SBSE/GC-MS. [8] The sulphur containing compounds consisted of dimethyl disulfide and 1,3-dimethyl trisulfide, which probably came from the degradation of sulfur-containing amino acids. [5,8,10,12,15,24,25] PLSR modelling correlation between sensory attributes and volatile compounds of liquors Partial least squares regression (PLSR) was adopted to draw the relationships between the sensory attributes and aroma compounds. PLSR was initially used to investigate the correlation between Chinese liquors, aroma compounds and sensory attributes. PLSR provided a two-factor model explaining 55% of the X-variance (aroma compounds) and 50% of the Y-variance (sensory attributes). Figure 2a showed the PLSR scores plot for the Chinese liquors based on the aroma compounds and sensory analyses. As shown in Figure 2b, it exhibited the resultant correlation loading plot. The inner ellipse and the outer ellipse indicated 50% and 100% of the explained variance, respectively. The sensory attributes and aroma compounds between the two ellipses indicated that they were well explained by the PLSR model. [26] The variables marked with small wileyonlinelibrary.com/journal/ffj

5 Characterization of chinese liquors by GC MS and sensory evaluation Figure 1. a) Graph of the mean sensory scores of the three aroma type liquors studied. Notations *, ** and *** indicate significance at p < 0.05, p < and p < 0.001, respectively. b) PCA bi-plot of sensory attributes for the three aroma type liquors (codesst1-st7, Li1-Li5, Sa1-Sa2, referring to the order listed in Section 2.1) circles were determined to be significant. The general structure revealed that three aroma type liquors were clearly separated from each other. Likewise, in Figure 2a, the same aroma type liquors clustered together while different aroma type liquors distributed into different groups. Sauce-aroma liquors were widely distributed at the positive values of PC1 and PC2, which could be attributed to the rich diversities of different commercial sauce-aroma liquors. In Figure 2b, four sensory attributes (caramel, fruity, sweet and grain) located inside the inner ellipse were poorly associated with aroma compounds. Cellar, fen, vinegar and sauce were located between the inner and outer ellipse. These four sensory attributes were significantly 221 wileyonlinelibrary.com/journal/ffj

6 Z. Xiao et al. 222 Table 2. The identification and quantitation of volatile compounds (mg L 1 ) of Chinese liquors code RI A RI D compound identification B strong-aroma liquor light-aroma liquor sauce-aroma liquor P Regression Curves R 2 (n=7x3) (n=5x3) (n=4x3) average C SD average SD average SD Esters ethyl acetate Y= X MS, RI, Std 5.39b a a e ethyl propanoate Y= X MS, RI, Std 0.30b a a f ethyl 2-methylpropanoate Y= X MS, RI, Std 0.44b b a propyl acetate Y= X MS, Std nd 0.24a a f 2-methylpropyl acetate Y= X MS, RI, Std nd 0.26a a e ethyl butanoate Y= X MS, RI, Std 19.24a b a f ethyl 2-methylbutanoate Y= X MS, RI, Std 0.54ab b a f ethyl 3-methylbutanoate Y= X MS, RI, Std 0.84b b a isoamyl acetate Y= X MS, RI, Std nd ethyl pentanoate Y= X MS, RI, Std 13.79a b a pentyl acetate Y= X MS, Std nd 0.02b a butyl butanoate Y= X MS, RI, Std nd 0.02a a g ethyl hexanoate Y= X MS, RI, Std a b b y isoamyl butyrate Y= X MS, RI, Std 0.31a b a hexyl acetate Y= 3.29E X MS, RI, Std 0.15a a a k propyl hexanoate Y= X MS, RI,Std 0.68a b a f ethyl heptanoate Y= X MS, RI, Std 8.15a b ab ethyl lactate Y= X MS, RI, Std 2.30a a a ethyl 2-hexenoate hexyl acetate E MS(97, 99, 55) F nd 0.00 nd h 2-methylpropyl hexanoate hexyl acetate E MS, RI 0.29a 0.26 nd a h isoamylpentanoate hexyl acetate E MS, RI 0.10a 0.07 nd a ethyl 2-hydroxy-3- ethyl decanoate E MS(73, 32) 0.14a a a methylbutanoate ethyl octanoate Y= X MS, RI, Std 2.61a a a h isoamylhexanoate ethyl decanoate E MS, RI 1.32a b b g ethyl nonanoate Y= X MS, RI,Std 0.09a a a y isoamyl lactate hexyl acetate E MS, RI 0.06a a ab k hexyl hexanoate Y= X MS, RI, Std 0.35a b b ethyl decanoate Y= X MS, RI, Std 0.10b a b 0.15 < h ethyl trans-4-decenoate Y= X MS, RI,Std 0.01b b a diethyl butanedioate Y= X MS, RI, Std 0.12b a b g ethyl benzeneacetate Y= X MS, RI, Std 0.34b b a phenethyl acetate Y= X MS, RI, Std 0.02b a ab ethyl 3-hydroxyhexanoate ethyl decanoate E MS (32,117) a 0.03 nd e ethyl benzenepropanoate ethyl benzeneacetate E MS, RI 0.94a a a e diethyl suberate ethyl decanoate E MS, RI b a ethyl oleate ethyl decanoate E MS (55, 88, 265) 0.67a b a wileyonlinelibrary.com/journal/ffj

7 Characterization of chinese liquors by GC MS and sensory evaluation Table 2. (Continued) code RI A RI D compound identification B strong-aroma liquor light-aroma liquor sauce-aroma liquor P Regression Curves R 2 (n=7x3) (n=5x3) (n=4x3) average C SD average SD average SD ethyl linoleate ethyl decanoate E MS (67, 81, 95) 1.11a b a Alcohols butanol Y= X MS, RI, Std 0.09a a a pentanol Y= X MS, RI, Std 0.17a 0.22 nd 0.06a butanol Y= X MS, RI, Std nd 0.03a a methyl-1-butanol Y= X MS, RI, Std nd 0.04a a methyl-1-butanol Y= X MS, RI, Std 1.67b a b pentanol Y= X MS, RI, Std nd 0.03a a e 2-heptanol Y= X MS, RI, Std 0.19a a a g 1-hexanol Y= X MS, RI, Std 1.30a a a k 3-octanol Y= X MS, RI, Std nd 0.02b a r 1-octen-3-ol Y= X MS, RI, Std 0.03b a ab g 1-heptanol Y= X MS, RI, Std 0.18a a a f 2-ethyl-1-hexanol Y= X MS, RI, Std 0.06b b a e 2-nonanol Y= X MS, RI, Std nd 0.05a a h 2,6-dimethyl-4- Y= X MS, RI, Std 0.44a a a heptanol octanol Y= X MS, RI, Std 0.28a a a k 1-nonanol Y= X MS, RI, Std 0.09b ab a (3Z)-3-nonen-1-ol 1-octen-3-ol E MS (55, 68, 81) nd a a undecanol Y= X MS, Std nd nd benzyl alcohol Y= X MS, RI, Std nd nd phenethyl alcohol Y= X MS, RI, Std 1.60c b a 3.27 < dodecanol Y= X MS, RI, Std nd nd Acids acetic acid Y= X MS, RI, Std nd 0.26a a butanoic acid Y= X MS, RI, Std 0.36b c a 0.28 < methyl butanoic acid Y= X MS, RI, Std 0.20b 0.17 nd 0.45a k pentanoic acid Y= X MS, RI, Std 0.36a 0.18 nd 0.30a g hexanoic acid Y= X MS, RI, Std 26.41a b b 2.48 < k heptanoic acid Y= X MS, RI, Std 1.78a b ab k octanoic acid Y= X MS, RI, Std 4.45b a ab k nonanoic acid Y= X MS, RI, Std 0.35b ab a g decanoic acid Y= X MS, RI, Std 0.50b a ab f dodecanoic acid Y= X MS, RI, Std nd nd Actals ,1-diethoxy ethane Y= X MS, RI, Std 0.27a a 0.00a 0.65a y 1,1-diethoxy hexane E MS, RI 0.20b b a wileyonlinelibrary.com/journal/ffj

8 Z. Xiao et al. 224 Table 2. (Continued) code RI A RI D compound identification B strong-aroma liquor light-aroma liquor sauce-aroma liquor P Regression Curves R 2 (n=7x3) (n=5x3) (n=4x3) average C SD average SD average SD 1,1-diethoxy-3- methylbutane h 1-(1-ethoxyethoxy) -pentane 1,1-diethoxy hexane E MS, RI nd b a ,1-diethoxy hexane Y= X MS, Std 0.05b ab a ,1,3,3-tetraethoxypropane Y= X MS, Std 0.25b ab a g 1,1,3-triethoxy propane 1,1-diethoxy hexane E MS, RI 0.20a a a h 1,1-diethoxy nonane Y= X MS, Std 0.04a a 0.01 nd (2,2-diethoxyethyl) -benzene 1,1-diethoxy hexane E MS(103, 91, 75) 4.50b a a Aldehydes methylbutanal Y= X MS, RI, Std nd 0.26a a hexanal Y= X MS, RI, Std nd 0.10a a f heptanal Y= X MS, RI, Std nd nd k octanal Y= X MS, RI, Std nd 0.04a a nananal Y= X MS, RI, Std 0.11a a a decanal Y= X MS, RI, Std 0.05a a 0.02 nd i benzaldehyde Y= X MS, RI, Std 2.07b 0.06 nd 21.02a (2E)-2-nonenal Y= X MS, RI, Std nd 0.07a a h (E,E)-2,4-decadienal Y= X MS, RI, Std 0.03a a 0.13 nd Ketones f 2-heptanone Y= X MS, RI, Std nd 0.04b a f 3-octanone Y= X MS, RI, Std nd 0.04a a octanone Y= X MS, RI, Std 0.45a a a f 6-methyl-5-hepten- Y= X MS, RI, Std nd nd one g 2-nonanone Y= X MS, RI, Std 1.69a a a k 2-decanone Y= X MS, RI,Std 0.19a a 0.01 nd f 2-undecanone Y= X MS, RI, Std 0.10a a a (2,6,6-trimethyl-1,3-0.02a a a cyclohexadien-1-yl) -2-buten-1-one (5E)-6,10-dimethyl-5,9- undecadien-2-one Phenols g 2-methoxy-4-methyl phenol 2-heptanone E MS (69, 121, 190) 2-heptanone E MS (43, 69, 151) 0.01a a a Y= X MS, RI, Std 0.45a a 0.23 nd g phenol Y= X MS, RI, Std nd nd wileyonlinelibrary.com/journal/ffj

9 Characterization of chinese liquors by GC MS and sensory evaluation Table 2. (Continued) code RI A RI D compound identification B strong-aroma liquor light-aroma liquor sauce-aroma liquor P Regression Curves R 2 (n=7x3) (n=5x3) (n=4x3) average C SD average SD average SD k 4-ethyl-2-methoxy phenol Y= X MS, RI, Std 1.28a a nd g 4-methyl phenol Y= X MS, RI, Std 0.73a 0.58 nd 0.05b k 4-ethyl phenol Y= X MS, RI, Std 0.12a a 1.53 nd thymol Y= X MS, Std 0.54b a ab Pyrazines k trimethyl pyrazine Y= X MS, RI, Std 0.05b b a k tetramethyl pyrazine Y= X MS, RI, Std nd 1.36a a h 2,5-dimethyl-3-n- trimethyl pyrazine E MS, RI nd 0.00 nd pentylpyrazine butyl-3,5-dimethyl pyrazine trimethyl pyrazine E MS (122) nd 0.00 nd Furans y 2-n-butyl furan Y= X MS, RI, Std 5.97b b a k furfrual Y= X MS, RI, Std 1.39b b a 5.6 < h 5-methyl-2-furfural Y=8.27E X MS, RI, Std nd nd y 3,4-dihydro-2H-1-2-furfual E MS, RI nd a a benzopyran-2-one dihydro-5-pentyl-2 2-furfual E MS (85) nd nd (3H)-furanone Sulfur-containing compounds dimethyl disulfide Y= X MS, RI, Std nd nd ,3-dimethyl trisulfide Y= X MS, RI, Std 0.05b b a 0.08 < A Retention Indices calculated of unknown compounds on a HP-INNOWAX capillary column ( mm 0.25μm) with a homologous series of n-alkanes (C7-C30). B Identification based on MS (mass spectrometry) or RI (retention index) or Std (reference compounds). MS means the volatile shows ion fragments matched to those in the Mass Spectral Library(NIST08, Wiley7n. l). RI means RI of the volatiles agree with literature data or flavornet database. Std means the volatiles were detected to have their ion fragments and RIs matched to those of reference compounds; C Values are mean ± standard deviation (mg L 1 ), calculated through an internal standard method; ANOVA tests were carried out between types of liquors, different letters (a c) within each row indicate significantly different between mean concentrations for each compound, according to Duncan s multiple range tests (p< 0.05). D From the flavornet database ( Acree, 2004 (on C20M stationary phase); in the literature (e: Jorge A. Pino& Sebastian Tolle, 2012); [23] (f: Giri, Anupam& Osako, Kazufumi, 2010); [24] ( g: Gao W, Fan W, Xu Y, 2014); [14] k:(fanw,shenh,xuy,2011); [8] y: (Xiao Z, Yu D, Niu Y, 2014); [10] h: (Niu Y, Yu D, Xiao Z, 2014); [25] i:(niu Y, Zhang X, Xiao Z, 2011). [26] E For the compounds with no available standard, the concentration of the volatile compounds in Chinese liquors was determined by using the calibration curves of a similar functional group standards, the respective standard was marked out. F In the case only MS allows the identification, the main ions ware given. nd: not detected. 225 wileyonlinelibrary.com/journal/ffj

10 Z. Xiao et al. Figure 2. a) An overview of the variation found in the mean data from partial least squares regression (PLSR) scores plot for the Chinese liquors based on the GC-MS and sensory analyses. Chinese liquors of St1-St7, Li1-Li5 and Sa1-Sas referring to the code liquors listed in Section 2.1. b) An overview of the variation found in the mean data from partial least squares regression (PLSR) correlation loadings plot for aroma compounds (X-matrix) and sensory attributes (Y-matrix). Codes referring to aroma compounds listed in Table correlated with aroma compounds. Cellar, located on the left side of PC1, which was the primary aroma of strong-aroma liquors, was positively and significantly correlated with ethyl hexanoate (r=0.61), hexanoic acid (r=0.74), hexyl hexanoate (r=0.71) 4-methyl phenol (r=0.64), ethyl pentanoate (r=0.58), isoamyl hexanoate (r=0.65), pentanoic acid (r=0.64), octanoic acid (r=0.50). On the right side of PC1, it could be noted that vinegar were positively and significantly connected with ethyl acetate (r=0.80), 2-methylpropyl acetate (r=0.70), isoamyl acetate (r=0.68), isoamyl lactate (r=0.67), ethyl decanoate(r=0.71), diethyl butanedioate (r=0.73), 2-phenethyl acetate (r=0.72), 3,4-dihydro-2H-1-benzopyran-2-one (r=0.74) while fen, which was supposed to be the primary aroma of light-aroma liquors, was mainly correlated with 1-dodecanol (r=0.77) and ethyl decanoate (r=0.80), 3-methyl-1-butanol (r=0.60), thymol (r=0.66). Sauce, which was the primary aroma of sauce-aroma liquors and located at the positive values of PC1 and PC2, had significant and positive connotation with ethyl 2-methylpropanoate (r=0.74), ethyl 3-methylbutanoate (r=0.70), furfural (r=0.91), 1,3-dimethyl trisulfide (r=0.92), trimethyl pyrazine (r=0.71), ethyl benzeneacetate (r=0.72), benzaldehyde (r=0.72), 1-octanol (r=0.56). Conclusions In conclusion, sensory descriptive analysis and SBSE/GC-MS were implemented for characterization of three typical aroma Chinese liquors. Different aroma liquors possessed different aroma profiles. Strong-aroma liquors were mainly characterized by cellar aroma; light-aroma liquors presented the highest scores in fen, floral, grain, sweet and vinegar aromas while sauce-aroma liquors were mainly characterized by sauce, fruity and caramel aromas. Results of PLSR indicated that ethyl hexanoate, hexanoic acid, hexyl hexanoate, 4-methyl pheno, ethyl pentanoate, isoamyl hexanoate and pentanoic acid were significantly and positively associated with cellar aroma and strong-aroma liquors. Vinegar and fen aroma were strongly linked with light-aroma liquor and associated with of ethyl acetate, 2-methylpropyl acetate, isoamyl acetate, isoamyl lactate, wileyonlinelibrary.com/journal/ffj

11 227 Characterization of chinese liquors by GC MS and sensory evaluation ethyl decanoate, diethyl butanedioate, 2-phenethyl acetate, 3,4-dihydro-2H-1-benzopyran-2-one 1-dodecanol, ethyl decanoate, 3-methyl-1-butanol and thymol. While sauce-aroma liquors were positively correlated with caramel, fruity and sauce sensory descriptors and volatile compounds such as ethyl 2-methylpropanoate, ethyl 3-methylbutanoate, furfural, 1,3-dimethyl trisulfide, trimethyl pyrazine, ethyl benzeneacetate, benzaldehyde and 1-octanol. In summary, it is concluded that different aroma type liquors can be significantly discriminated by their volatile compounds and sensory attributes. The investigation is very useful for improvement of characteristic aroma of different aroma-types of Chinese liquors. Acknowledgements The research was supported by the National Youth Science Foundation of China (No ), The National Natural Science Foundation of China (No ), The National Key Technology R&D Program (2011BAD23B01) and Shanghai Engineering Technology Research Center of Fragrance and Flavor (15DZ ). Dan Yu, Yunwei Niu, and Zuobing Xiao have has received research grants from Institution A. Compliance with ethics requirements This research has fully complied with research ethics. Conflict of interest Ning Ma declares that he has no conflict of interest. Jiancai Zhu declares that he has no conflict of interest. This article does not contain any studies with human or animal subjects. References 1. C.Wang,D.Shi,G.Gong.World Journal of Microbiology and Biotechnology. 2008, 24, C. Zhang, C. Shen, Z. Ao, W. Tao, W. Chui, S. Zhang. Eur. Food Res. Technol. 2012, 234, H. Li, W. Huang, C. Shen, B. Yi. Food Bioprod. Process. 2012, 90, W. Fan, M. C. Qian. J. Agric. Food Chem. 2005, 53, W. Fan, M. C. Qian. J. Agric. Food Chem. 2006, 54, W. Fan, M. C. Qian. Flavour Frag. J. 2006, 21, S. Zhu, X. Lu, K. Ji, K. Guo,Y. Li, C. Wu, G. Xu.Anal. Chim. Acta. 2007, 597, W. Fan, H. Shen, Y. Xu. J. Sci. Food Agric. 2011, 91, M. P. Saenz-Navajas, Y. S. Tao, M. Dizy, V. Ferreira, P. Fernandez- Zurbano. J. Agric. Food Chem. 2010, 58, Z. Xiao, D. Yu, Y. Niu, F. Chen, S. Song, J. Zhu, G. Zhu. J. Chromatogr. B. Analyt. Technol. Biomed. Life Sci. 2014, 92, H.Sun,H.Ni,Y.Yang,F.Chen,H.Cai.A. Xiao. Flavour Frag. J. 2014, 29, Y. Niu, D. Yu, Z. Xiao, J. Zhu, S. Song, G. Zhu. Food Analytical Methods. 2015, 8, S. Wold, M. Sjöström, L. Eriksson. Chemometr. Intell. Lab. 2001, 58, R. Noguerol-Pato, M. Gonzalez-Alvarez, C. Gonzalez-Barreiro, B. Cancho- Grande. J. Simal-Gandara. Food Chem. 2013, 139, W. Gao, W. Fan, Y. Xu. J. Agric. Food Chem. 2014, 62, R. M. González-Rodríguez, R. Noguerol-Pato, C. González-Barreiro, B. Cancho-Grande, J. Simal-Gándara. Food Res. Int. 2011, 44, R. Noguerol-Pato, M. Gonzalez-Alvarez, C. Gonzalez-Barreiro, B. Cancho-Grande, J. Simal-Gandara. Food Chem. 2012, 134, M. Gonzalez Alvarez, C. Gonzalez-Barreiro, B. Cancho-Grande, J. Simal- Gandara. Food Chem. 2011, 129, R. Noguerol-Pato, C. González-Barreiro, B. Cancho-Grande, J. Simal- Gándara. Food Chem. 2009, 117, C. Flanzy. Enología: Fundamentos científicos y tecnológicos, 2nd edn. AMV Ediciones: Ediciones Mundi-Prensa, Madrid, T. Shinohara. Agricultural and Biological Chemistry. 1985, 49, W. Fan, Y. Xu, Y. Zhang. J. Agric. Food Chem. 2007, 55, P. Ribéreau-Gayon, Y. Glories, A. Maujean, D. Dubourdieu. Handbook of Enology: The Chemistry of Wine Stabilization and Treatments Jon Wiley & Sons, Chichester, England, 2nd edn, J. A. Pino, S. Tolle, R. Gök, P. Winterhalter. Food Chem. 2012, 132, A. Giri, K. Osako, T. Ohshima. Food Chem. 2010, 120, Y. Niu, X. Zhang, Z. Xiao, S. Song, K. Eric, C. Jia, H. Yu, J. Zhu. J. Chromatogr. B. Analyt. Technol. Biomed. Life Sci. 2011, 879, wileyonlinelibrary.com/journal/ffj