Unraveling the acetals as ageing markers of Chinese Highland Qingke Baijiu using comprehensive two-dimensional gas chromatography–time-of-flight mass spectrometry combined with metabolomics approach


 
 
 The ageing process has a significant impact on the aroma of Chinese Baijiu, which could strengthen the desirable flavor characteristics and reduce the undesirable ones. The aim of this study was to observe the initiation of meaningful changes in volatile fraction and locate the ageing markers during ageing storage of Chinese Highland Qingke Baijiu.
 
 
 
 Samples of Chinese Qingke Baijiu were aged for 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, and 11 months before analysis. The samples were isolated by liquid–liquid extraction and then analyzed by comprehensive two-dimensional gas chromatography–time-of-flight mass spectrometry. The acquired data were processed by untargeted and targeted metabolomics approach to locate the ageing markers.
 
 
 
 The untargeted metabolomics analysis (hierarchical clustering analysis, HCA) shows that the chemical composition of Qingke Baijiu presents a statistically significant deviation from the reference scenario after 5 months. Subsequently, supervised statistics analysis (orthogonal partial least squares discrimination analysis) was performed to locate the markers, which changed significantly during ageing. Fifteen markers were located, and seven of them were acetals. Notably, 1,1-diethoxy-propane, 1,1-diethoxy-butane, and 1,1-diethoxy-3-methyl-butane are important contributors to the flavor of Chinese Baijiu. The identified markers were applied for the untargeted metabolomics (HCA), and the results revealed that these markers could divide the Qingke Baijiu into two ageing stages, 0–5 months and 6–11 months.
 
 
 
 The results suggest that it is a valuable tool for monitoring the changes of volatile compounds and locating the age markers in Chinese Baijiu.



Introduction
Baijiu, or Chinese liquor, is a type of very traditional distilled spirits with a long history in China. The production process of Baijiu involves saccharification, fermentation, distillation, ageing, and blending (Zheng et al., 2017;Liu and Sun, 2018). Chinese Baijiu is commonly made from grain as raw material and uses Jiuqu as fermentation starter. Jiuqu is not only a starter culture, but also plays an important role in the aroma formation during fermentation Wang et al., 2020). Interestingly, Chinese Baijiu could be classified into 12 aroma types, among them strong, soy sauce, rice and light aroma are the four basic aroma types (Jin et al., 2017;Liu and Sun, 2018). Highland Qingke Baijiu is one of the representative brands of light aroma type Baijiu. It has an enjoyable floral and fruity aroma, mellow sweetness, and is refreshing with a pure highland barley aroma . Unlike other Chinese Baijiu, Qingke Baijiu is unique because of using hull-less highland barley as the raw material and being produced in Qinghai-Tibetan Plateau with an average altitude of 4000 m . The fresh raw Qingke Baijiu has unwanted aromas, such as 'green' and 'raw' aromas (Zhu et al., 2016). These undesirable aromas could be due to the low molecular mass volatile compounds, including methanol, sulfides, acetaldehyde, etc. It has been reported that the ageing process has a significant impact on the aroma of Chinese Baijiu, which could strengthen the desirable flavor characteristics and reduce the undesirable ones (Zhu et al., 2020). However, very few studies have focused on the changes in the volatile composition of Qingke Baijiu.
For many food matrices including wines, cheeses, and beverages, the flavor can be improved during the ageing process, which could be attributed to numerous reactions (Song et al., 2018;Huang et al., 2020). To build a strategy to better control the flavor quality during storage, it is vital to understand the changing trends of volatile compounds (Pozo-Bayón et al., 2009). For this purpose, some volatile components have been regarded as potential ageing markers, such as lactones, furans, phenols, and acetals (Zhu et al., 2016). Gas chromatography-mass spectrometry (GC-MS) has been commonly used to trace the changes of volatile compounds for many foods (Gao et al., 2018(Gao et al., , 2021Jia W. et al., 2020). 1,1-Diethoxymethane and methanethiol were analyzed by GC-MS and it was demonstrated that these two compounds could be used as ageing markers of roasted sesame-like flavor type Baijiu (Zhu et al., 2016). Besides, the changes of key odorants and flavor profiles of Chinese Laobaigan Laowuzeng Baijiu during its one-year ageing were evaluated by GC-MS and gas chromatography-olfactometry (GC-O) (Zhu et al., 2020), despite GC-MS technology being successfully applied in the analysis of food products (Pozo-Bayón et al., 2009). However, GC-MS lacks coeluting compounds and can overlook some crucial changes in markers. Instead, comprehensive two-dimensionalgas chromatography (GC×GC) is the most powerful multidimensional separation tool for digging deeply into volatile components in complex samples (Nunes et al., 2017;Song et al., 2019;Liu J. W. et al., 2020;Xiang et al., 2020;Song et al., 2021). Moreover, comprehensive twodimensional gas chromatography-time-of-flight mass spectrometry (GC×GC-TOFMS) can maintain useful chemical information, particularly the sample's volatile compounds (Pozo-Bayón et al., 2009). However, the analysis of GC×GC generates enormous and multidimensional data, thus interpreting and assessing the information about volatiles is a challenging task (Magagna et al., 2016). A metabolomics approach can efficiently resolve this problem. This powerful tool, including hierarchical clustering analysis (HCA), principle component analysis, partial least squares regression, and orthogonal partial least squares discrimination analysis (OPLS-DA), has been extensively applied in the study of food matrices (Pang et al., 2019;Liu S. Q. et al., 2020;Pu et al., 2021;Yang et al., 2021). The GC×GC-TOFMS, combined with a metabolomics approach, has been successfully applied in quality control, regional discrimination, technological treatment, storage conditions, and differentiation (Jia X. et al., 2020;Song et al., 2020a;Song et al., 2020b). To the best of our knowledge, most of the studies in Qingke Baijiu focused on the identification and quantitation of aroma compounds . Very few studies learned the changes of volatiles and located the ageing markers of Qingke Baijiu during ageing storage.
In the present study: (1) The volatile composition was analyzed by liquid-liquid extraction combined with GC×GC-TOFMS; (2) The unsupervised statistical methods were applied to reveal the initiation of meaningful changes in volatile fraction during ageing storage; (3) The untargeted metabolomics analysis was used to locate the markers that change significantly during ageing; (4) The targeted metabolomics analysis was performed to verify the capability of the ageing markers for the differentiation of the Qingke Baijiu in the different ageing periods.

Samples
Samples were obtained from Qinghai Highland Barley Wine Co. Ltd. (Huzhu, China). The samples were aged for 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, and 11 months at the time of their analysis with an average ethanol content of 65.5 per cent (v/v). The Qingke Baijiu samples were produced using the same manufacturing technique and raw materials.

Isolation of volatile compounds by liquid-liquid extraction
A total of 25 mL of Qingke Baijiu sample spiked with 20 μL of naphthalene-D8 (200 mg/L) was diluted to the ethanol content of 10 per cent (v/v). The diluted sample was saturated by adding the sodium chloride (46 g), and then extracted by dichloromethane three times (30 mL per extraction). The collected organic extraction was first dried and subsequently enriched to 0.5 mL using a vigreux column. The final obtained enriched Qingke Baijiu samples were stored at -40 °C for analysis (Song et al., 2020a;Song et al., 2020b).

GC×GC-TOFMS
The GC×GC-TOFMS instrument (Pegasus 4D, Leco, St. Joseph, MI, USA) system was executed on a 7890B gas chromatograph and a TOFMS detector. The gas chromatography (GC) had two ovens and was equipped with a dual-stage quad-jet thermal modulator. The type of the injector was programmed temperature vaporizer, which was set at a temperature of 250 °C and kept in splitless mode. The first dimensional column was a polar column: DB-WAX column (30 m×0.25 mm, 0.25 μm; Agilent, Palo Alto, CA, USA), and the second dimensional column was a nonpolar column: DB-5 column (2 m×0.1 mm, 0.1 μm; Agilent, Palo Alto, CA, USA). Helium was used as the carrier gas and kept at a flow of 1 mL/min. The oven temperature was set as follows: The initial temperature was set as 50 °C and kept for 2 min. Subsequently, the temperature increased to 150 °C at a rate of 3 °C/min, then increased to 230 °C at 5 °C/min, and finally kept for 10 min. The time for hot pulse and cold pulse were 1 s and 1.5 s, respectively. The acquisition rate, the voltage and the ionization energy for the detector were 100 Hz, 1450 V, and 70 eV, respectively (Song et al., 2020a;Song et al., 2020b).

Data processing
The Qingke Baijiu samples, which including 35 observations (11 samples in three replicates and one sample in two replicates) were deconvoluted and aligned by statistical software (Leco, St. Joseph, MI, USA). The alignment of the peaks for all the observations was according to their first-and second-dimensional retention times and the masses. The volatile compounds that appeared under 50 per cent of the observations were taken out (Song et al., 2020a). The data were pretreated by unit variance scaling and logarithmic transformation and then subjected to HCA, which was used to visualize meaningful trends in the samples. Subsequently, the data were applied to the OPLS-DA for locating the markers, which changed significantly during the ageing process. The model parameters were validated by k-fold cross-validation (k-fold-CV). The proportion of the model's variance was explained by R 2 Y , and the predictive ability was shown by Q 2 Y . The potential relationship among the located markers and the samples was elucidated by heatmap analysis. The significant differences (P values) and the fold-change were determined by the t-test. The OPLS-DA were performed by SIMCA-p version 14.0 (Umetrics, Umea, Sweden). The HCA and heatmap analysis were carried out by MetaboAnalyst (https://www. metaboanalyst.ca/; Song et al., 2020a).

Results and Discussion
Volatile compounds determined in Qingke Baijiu by GC×GC-TOFMS According to the literature search, the volatile compounds in Qingke Baijiu during the ageing process were analyzed for the first time to identify the ageing markers. A total of 282 volatile compounds in 35 observations (11 Baijiu samples and their three analytical replicates and one Baijiu sample repeated twice) have been identified based on chromatographic analyses (Figure 1). Subsequently, the information about compounds was used to develop the data matrix. The untargeted metabolomics analysis (HCA and OPLS-DA) for Qingke Baijiu To better get a preliminary sight of the objective clustering of the Qingke Baijiu samples according to the ageing process, HCA was conducted on all 35 observations. The obtained dendrogram is shown in Figure 2. The squared Euclidian distance measured the proximity between two observations, and Ward's method was taken as a linkage rule. It is suggested that the ages of the samples present were a significant factor of variation among the different samples. In Figure 1, a clear separation was observed between two large groups; one group included the samples with an ageing time of 0-5 months, and the other group had the samples with an ageing time of 6-11 months. For each main group, different subgroups could be observed, corresponding to samples with different ageing times. There was an interesting observation to note that the repetitions of the same sample were mostly correctly clustered, except one for the 2-month-old and 3-month-old samples, which indicated that the reproducibility of the analytical method was excellent.
Then, the supervised statistics analysis, OPLS-DA, was conducted to locate ageing markers of Qingke Baijiu. Firstly, the built OPLS-DA model was measured by the k-fold-CV (k=7). The obtained results revealed that R 2 Y (the explained variation) and Q 2 Y (the predictive capability) for the OPLS DA model were 0.986 and 0.973, respectively, demonstrating that the built OPLS-DA model was distinguished due to the Q 2 Y being closer to 1. From Figure 3A, the samples from two ageing times were clearly distinguished in the OPLS-DA score plot. Moreover, permutation tests (n=200) were carried out to assess the fitting degree of the OPLS-DA model. The permutation tests randomly rearranged the experiments by changing the sort order of the classification variables (Y) and randomly assigned Q 2 Y up to 200 times (Song et al., 2020a). According to the previous study, the R 2 Y intercept=~0.3 and Q 2 Y intercept<0.05 revealed that the model was a good fit (Song et al., 2020a). In the present study, the intercept value of Q 2 Y for the permutation test was 0.39, which demonstrated that the models were not overfitting and met the analysis requirements. The R 2 and Q 2 values of the permutation test indicated that the initial model outperformed the randomly permuted models ( Figure 3B) (Song et al., 2020a).

Heatmap analysis of markers and changes of markers during ageing
To illustrate the relationship between the identified marker compounds, the relative concentrations of the identified markers in the samples of 0-5 months and 6-11 months of Chinese Qingke Baijiu were conducted to a heatmap analysis, and the color (from light to dark) demonstrated the relative intensity change from low to high. The differences between the marker compounds among the samples are shown in Figure 4. The detailed information of the relative concentration of the identified markers in the 0-, 3-, 6-, 9-, and 11-month Qingke Baijiu was shown in Table 2. The HCA clearly classified the quantitated markers into two clusters. Cluster 1 consists of the samples of 0-5 months of Qingke Baijiu with significantly higher (P<0.01) concentrations, which included acetophenone, butanedioic acid, ethyl-3-methylbutyl ester, and dihydro-4-hydroxy-2(3H)-furanone.  GC-GC-TOFMS, comprehensive two-dimensional gas chromatography-time-of-flight mass spectrometry; MS, mass spectral; RI, retention index data; S, standard substance. Table 2. The relative concentration of the 12 identified markers in 0-, 3-, 6-, 9-, and 11-month Qingke Baijiu

No. Compound
Relative concentration (μg/L) 0-month 3-month 6-month 9-month 11-month  Acetophenone (sweet, fruity, floral) has been characterized as an important odorant in Chinese soy sauce aroma type Baijiu (Fan et al., 2012). More importantly, it has been regarded as an important difference between young and aged Huangjiu (rice wine). The flavor dilution (FD) factors of acetophenone were 16 times higher in the Huangjiu than in the young Huangjiu . As profiled in Figure 5, nine marker compounds showed a significantly higher concentration in 6-11 months than 0-5 months Chinese Qingke Baijiu. Except for diethyl carbonate, eight of them are acetals. 1,1-Diethoxypropane (fruity) has been reported as an important aroma contributor to Chinese strong and soy sauce aroma type Baijiu (Fan and Qian, 2006;Fan et al., 2012). Similarly, 1,1-diethoxy-3-methyl-butane also has a significant impact on the aroma of Chinese strong and soy sauce aroma type Baijiu, especially on aged strong aroma type Baijiu. Additionally, it contributed significantly to the flavor of Cognac wine (Fan and Qian, 2005;Fan et al., 2006Fan et al., , 2012Thibaud et al., 2019). 1,1-Diethoxy-butane also has been identified as an aroma-active compound in Cognac wine. For consideration of the statistical role of the acetals, 1,1-diethoxy-3-methyl-butane, 1-(1-ethoxyethoxy)-pentane, and 1,1-diethoxy-butane have been regarded as potential age markers of Madeira wines (Perestrelo et al., 2011).

Conclusion
In the present study, an efficient strategy to locate ageing markers of Chinese Qingke Baijiu was proposed by a metabolomics approach combined with GC×GC-TOFMS. Twelve marker compounds were accurately identified, and seven of them were acetals. In particular, 1,1-diethoxy-propane, 1,1-diethoxy-butane, and 1,1-diethoxy-3methyl-butane have been known as important contributors to the flavor of Chinese Baijiu. The findings are useful for the quality control of the ageing process of Chinese Qingke Baijiu. It appears that this strategy could be a useful tool for monitoring volatile changes during storage in Qingke Baijiu. Moreover, this strategy could be applied for the quality control, differentiation, and authenticity of Chinese Baijiu.

Supplementary Material
Supplementary material is available at Food Quality and Safety online.