Comparison of Volatiles in Different Jasmine Tea Grade Samples Using Electronic Nose and Automatic Thermal Desorption-Gas Chromatography-Mass Spectrometry Followed by Multivariate Statistical Analysis

Chinese jasmine tea is a type of flower-scented tea, which is produced by mixing green tea with the Jasminum sambac flower repeatedly. Both the total amount and composition of volatiles absorbed from the Jasminum sambac flower are mostly responsible for its sensory quality grade. This study aims to compare volatile organic compound (VOC) differences in authoritative jasmine tea grade samples. Automatic thermal desorption-gas-chromatography-mass spectrometry (ATD-GC-MS) and electronic nose (E-nose), followed by multivariate data analysis is conducted. Consequently, specific VOCs with a positive or negative correlation to the grades are screened out. Partial least squares-discriminant analysis (PLS-DA) and hierarchical cluster analysis (HCA) show a satisfactory discriminant effect on rank. It is intriguing to find that the E-nose is good at distinguishing the grade difference caused by VOC concentrations but is deficient in identifying essential aromas that attribute to the unique characteristics of excellent grade jasmine tea.


Introduction
It is a general belief that the pleasant aroma of the Jasminum sambac flower can relieve the mood of depression [1]. Moreover, the health effect of tea also has been widely confirmed [2]. Both of these concepts make jasmine tea a popular tea worldwide [3,4]. Traditionally, the processing of jasmine tea, includes the following seven steps (shown in Figure S1) of tea dhool preprocessing, fresh flowers maintenance, tea and flower combination, scenting, flower removal, drying, and packing [5]. Current Chinese National Standards subdivide jasmine tea into six grades according to the number of times of repeated scenting which affects the quality of both the Jasminum sambac flower and the tea dhool [6,7]. The floral fragrance adsorption and persistence are critical factors related to jasmine tea grading [8].

Automatic Thermal Desorption-Gas-Chromatography-Mass Spectrometry (ATD-GC-MS) Analysis
The volatile organic compound (VOC) of jasmine tea samples was analyzed using an ATD-GC-MS method, described by Zheng [24], with slight modification. A COLIN Tech Auto thermal desorption sampler (Chengdu Colin Analytical Technology Co., Ltd., Chengdu, China) and a Shimadzu 2010 gas-chromatography (GC) coupled with 8040 triple quadrupole mass spectrometry (TQ-MS) (Shimadzu Production Institute, Kyoto, Japan) was applied.

Extraction of Volatile Organic Compounds (VOC)
A QC-1S atmosphere sampling instrument (Beijing Kean Labor Insurance New Technology Co., Ltd., Beijing, China) was used for VOC extraction according to China's National Environmental Protection Standards [25]. The VOC analysis method was the same as Zheng et al. [24]. Briefly, 3.0 g of sample was weighted into a headspace bottle and ethyl decanoate (100 ppm, 15 µL) was added to the samples as the internal standard. Then, the headspace bottle was sealed and equilibrated at 55 • C for 20 min. Afterward, the sorbent tube (Chengdu Colin Analytical Technology Co., Ltd., Chengdu, China) was connected to the atmosphere sampling instrument and headspace bottle according to the flow direction of the sorbent tube with polytetrafluoroethylene (PTFE) pipes. Finally, volatile components were collected at 200 mL/min flow rate for 30 min. After sample collection, both ends of the sorbent tube were sealed with PTFE caps and transported to the laboratory for analysis.

Thermal Desorption
Thermal desorption was conducted by a COLIN Tech Auto thermal desorption sampler (Chengdu Colin Analytical Technology Co., Ltd., Chengdu, China). The primary thermal desorption of sampling tube was carried out at 250 • C for 5 min. To introduce trapped compounds into the gas chromatograph, the cold trap was then heated rapidly from −25 • C to 300 • C. The temperature of the valve and transfer line were maintained at 200 • C during analysis. Then, the whole system was baked at 300 • C for 3 min in preparation for the next sample analysis.

Gas-Chromatography-Mass Spectrometry Analysis
Volatile organic compounds were identified using a 2010 GC coupled with an 8040 TQ-MS system (Shimadzu Corporation, Kyoto, Japan). The capillary column was a Shimadzu Rtx-5MS capillary column (30 m × 0.25 mm × 0.25 µm), and the carrier gas was helium at 1.0 mL/min. The split ratio was 1:40. The inlet temperature was 240 • C. The gradient temperature program was as follows: initial oven temperature was 40 • C, held for 3 min; 40-120 • C at 5 • C/min, held for 5 min; 120-240 • C at 30 • C/min, held for 8 min. The ionization mode of the MS was electron impact (EI). The temperatures of the interface and ion sources were 280 • C and 230 • C, respectively. The acquisition mode was full scan.

Identification of Volatile Organic Compounds (VOC)
Shown in Table 1, the volatile compounds were identified by matching their mass spectra fragmentation patterns, retention index with those stored mass spectra libraries (NIST 11.L and Wiley 7), and combining them with existing works of literature [3,4,[26][27][28]. The relative content of identified compounds was obtained by comparing them with the peak area of internal standards ( Table 2).

Electronic nose (E-Nose) Measurements
An ISENSO iNose E-nose system (Shanghai Ongshen Intelligent Technology Co., Ltd., Shanghai, China) was used to profile volatile fingerprints. Shown in Table S2, the gas detectors of the E-nose system were composed of ten metal oxide sensors (MOS), each of which was sensitive to different volatile organic compounds [29], respectively.
A portion of each sample (3.0 g) was weighed into a headspace bottle (60 mL) and equilibrated in a 55 • C water bath for 40 min. Then, the gas in the headspace was pumped over the sensor surfaces for 5 min at a constant flow rate of 800 mL/min. Finally, cleaning the probe with continuously pumped filtered air until all sensors' baseline value returned to 1.00 was preparation for the next sample analysis. The stable value of each sensor was extracted for data processing.

Statistical Analysis
Soft independent modelling by class analogy (SIMCA) 14.1 (Umetrics AB, Umea, Sweden) was used for partial least squares-discriminant analysis (PLS-DA) and principal component analysis (PCA). The heatmap and hierarchical cluster analysis (HCA) was conducted using MetaboAnalyst web (https://www.metaboanalyst.ca/MetaboAnalyst/faces/home.xhtml). Statistical package for the social sciences (SPSS) 21.0 (IBM, Chicago, IL, USA) were applied for multivariate statistical analysis. The differences among six grades of jasmine tea samples were estimated through analysis of variance (ANOVA). Regarding rank correlation analysis, the correlation between the response and grade of each substance was analyzed, the compounds with both a positive linear correlation or a negative correlation were screened out, respectively.

Identification of Volatile Organic Compounds in the Jasmine Tea by Automatic Thermal Desorption-Gas-Chromatography-Mass Spectrometry Spectrometry
To investigate the aroma characteristics of tested Jasmine tea samples, their volatile compounds were subjected to ATD-GC-MS, and the average relative amounts of identified volatiles were compared. Typical total ion chromatograms (TICs) are presented in Figure S2.
A total of 18 samples, with six different grades (named 1G, 2G, 2G, 3G, 4G, 5G, and 6G) and three repeats per category, were subjected to investigation of their aroma characteristics. The identified VOC and their corresponding amounts (mean ± standard deviation) were summarized; their significant differences also were tested ( Table 2; Table S4).
A total of sixty-three VOCs were identified (Table 1), including thirteen alcohols, five aldehydes, nineteen esters, twenty-three hydrocarbons, two ketones, one nitrogen compound, and one phenolic.

Alcohols
There were thirteen kinds of identified alcohol in the jasmine tea grade samples. Among these identified alcohols, linalool and benzyl alcohol, which are abundant in jasmine flowers [4,27,30], accounted for 52.32% and 30.91% of the total content of alcohol, respectively.
Linalool, imparts a floral, fruity, and woody odor in jasmine tea, and benzyl alcohol provides a sweet, roasted, mild, fruity and citrus-like aroma, were contained in both the tea dhool and jasmine flowers [4,27,31]. Here, the relative content of 3-hexen-1-ol in alcohols was lower than both linalool and benzyl alcohol. Meanwhile, they were closely related to the sensory attributes of grassy and lettuce-like aromas [27,32]. Furthermore, among alcohols, there were some volatile compounds with a negative correlation to the grade of jasmine tea, including cyclopentanone, 1-hexanol, (Z)-Linalool oxide and (E)-Linalool oxide. These four volatile compounds are found in green tea, and existing studies show that cyclopentanone, (Z)-Linalool oxide and (E)-Linalool oxide are negatively correlated with the grade of green tea [16,32]. It also was reported that phenyl ethyl alcohol, α-Terpineol, and geraniol were all derived from jasmine flowers, having floral or sweet odor [3,4,33].  a-d Means ± SD followed by the same letter, within a row, are not significantly different (p > 0.05); e 1G, 2G, 3G, 4G, 5G, and 6G represent the standard sample for the grade of jasmine tea from high rank to low rank; f MI, method of identification; N.D., peak intensity lower than triple signal-to-noise.

Aldehydes
Five aldehydes, namely, benzaldehyde, decanal, hexanal, (E, E)-2,4-heptadienal and β-cyclocitral, were detected in all six grades of jasmine tea. Although aldehydes comprised 0.88% of the identified volatile organic compounds (VOC), they still contributed a lot to the aroma performance due to their low odor threshold [33]. Among aldehydes, benzaldehyde, which provided almond, sugar and burnt aroma notes, and decanal which supplied herbal, fatty and citrus aroma notes, were proven to play an essential role in aroma [4,26,27]. Here, all five aldehydes were negatively correlated with the grade of jasmine tea. Interestingly, these volatile compounds, which were harmful to the quality of jasmine tea, had been reported in green tea or originated from tea dhool [3,34,35]. Existing studies also demonstrated that hexanal and (E, E)-2,4-Heptadienal were negatively correlated with the grade of Japanese Matcha [34].

Esters
Nineteen esters were found in all grades of jasmine tea. They accounted for 63.47% of identified volatile organic compounds (VOC) and positively correlated with the grade. Benzyl acetate, (Z)-3-hexanol benzoate, methyl salicylate, and their predecessors have confirmed methyl anthranilate as the main volatile aroma components of jasmine tea, which was consistent with the results of this study [3,4,9]. Among them were benzyl acetate, having floral, fruity odor notes, and (Z)-3-hexanol benzoate, with green, spicy, woody notes while herbaceous ones have prominent aroma characteristics of jasmine flowers [26][27][28]. Methyl anthranilate was described as similar to a peachy, sweet, fruity grape-like fragrance originated from jasmine flowers [3,4,27]. Methyl salicylate was considered to be a sweet, spicy, minty, wintergreen-like odor, and recognized as a significant aroma compound of black tea [3,36]. It is noteworthy that most volatile compounds of esters were positively correlated with the grade of jasmine tea and came from Jasmine flowers.

Hydrocarbons
Twenty-two hydrocarbons were identified in the tea samples. Despite the large number, it had a limited contribution to the aroma of tea [32,37]. Among them, α-farnesene, having floral and herbaceous odor notes, was the most abundant and recognized as one of the vital aroma components in jasmine tea [3,4,9]. Furthermore, Myrcene, Germacrene D and α-Farnesene, and so forth, were positively correlated with the grade of jasmine tea and were reported to originate from jasmine flowers [3,30], while α-pinene and limonene were negatively correlated with the grade.

Ketones
Two ketones, namely 6-methyl-5-heptane-2-one and acetophenone, were identified. The 6-methyl-5-heptane-2-one was described as sweet, fruity, with orange odor notes, and previous studies confirmed that the compound showed an increasing trend in the processing of Oolong tea [9,38]. Regarding acetophenone, it was identified in Oolong tea [39], green tea [40], Pu'er tea [41] and Jasmine tea [27], but had little effect on the tea aroma. Moreover, there was a negative correlation between the relative content of 6-methyl-5-hapten-2-one and the quality grade.

Nitrogen Compound
The nitrogen compound detected in the tea samples was indole, which provided nutty, floral, mothball, and burnt aroma notes. It was known as one of the main aroma components of jasmine tea and was positively correlated with the grade [3,4,8].

Phenols
The phenol detected in the tea samples was eugenol. It may originate from the Jasmine Flower and be considered to be a clove-like spicy smell [8,9,26].

Hierarchical Clustering Analysis (HCA)
To present VOC differences among different grade samples, a heat-map of eighteen samples versus identified compounds was plotted (Figure 1). The red color in the plot represents a higher content than the mean value; the blue color represents a lower content than the corresponding mean value. The HCA also was performed to get a cluster pattern among the six different grades. These six grades were subdivided into two categories, which were a high-grade group (including 1G, 2G, and 3G) and a low-grade group (including 4G, 5G, and 6G). By comparing the color intensity variation across all samples, we found that some compounds changed correlationally according to grade quality reduction (Table S4).  Figure 1, the compounds marked with the blue frame, named A, indicated an increasing trend which correlated with the decline of grade. There was a total of twelve compounds, including five aldehydes (hexanal, decanal, β-Cyclocitral, benzaldehyde, (E,E)-2,4-Heptadienal), four alcohols (linalool oxide, 1-Hexanol, (Z)-Linalool oxide, cyclopentanol), two hydrocarbons (limonene, α-Pinene), and one ketone (5-Hepten-2-one), with fragrant characteristics such as fruity, floral, woody, green, sweet or more [4,26,27]. It was intriguing to find that most of these substances came from tea dhool [3,28]. Furthermore, according to existing research [16,34,35,42], most of them were negatively correlated to the quality of green tea.

Shown in
Concerning the compounds in the red frame, B, (Figure 1), distinct rules were existing between the high-grade group (1G, 2G,3G) and the low-grade group (4G,5G,6G). Regarding the low-grade group, indicated as frame B 1 , positive linear correlations were existing. The amounts of both β-cadinene and (Z)-3-Hexenyl acetate, for example, were far higher in the higher grade. While, for the high-grade group, indicated as frame B 2 , there was not a simple linear relationship between their contents to the grade. Take β-cadinene as an example, the highest grade was in 1G, followed by 3G and 2G, however, Z-β-ocimene content was the highest in 2G, then in 1G and 3G. The reason may be that, in addition to the requirement of the intensity of flower fragrance, it is also an essential requirement for them to maintain Z-β-ocimene content at a moderate proportion, which could make its aroma coordinated.
Furthermore, as shown in Figure 1, there was a total of twenty-four compounds in frame B, including twelve esters, ten hydrocarbons, one alcohol, and one nitrogenous. It also is intriguing to find that most of them were absorbed from the jasmine flower [3,28,30]. It is remarkable that the main aroma components of jasmine tea (linalool, (Z)-3-Hexenyl benzoate, methyl salicylate, (Z)-3-Hexenyl acetate, α-Farnesene and indole) [3,4,9] were not linearly related to the grade, but were obviously rich in the high-grade group (1G, 2G,3G).

Partial Least Square-Discriminant Analysis (PLS-DA)
A supervised PLS-DA was approached to investigate the differences among standard grade samples. Shown in Figure 2A, the scores of the principal component (PC) 1 (abscissa) and PC2 (ordinate) were new variables summarizing variables. The scores were orthogonal, which were completely independent of each other. The score of PC1 explains the largest variation of the X space, followed, by PC2. Hence, the scatter plot of PC1 versus PC2 was a window displaying how the X observations were situated concerning each other. Significant discrimination, according to the data matrix of the volatile compounds in the six grades, was observed. Two groups of tea samples with a higher grade (1G and 2G) were distributed in the fourth quadrant, three groups of tea samples with a lower grade (3G, 4G, 5G) were distributed in the first quadrant and the second quadrant, while the lowest group of tea samples (6G) were distributed in the third quadrant alone. The high grade explained the variance (R 2 Y = 0.966) and cross-validated predictive capability (Q 2 = 0.979), manifesting the model's feasibility.   Figure 2C, displays the relation between the X-variables and the Y-variables. Moreover, X-variables situated in the vicinity of the dummy Y-variables have the highest discriminatory power between the classes. Striking was that the plot in Figure 2C further explains the six grades of jasmine tea samples for differences in specific volatile components. Shown in Figure 2D, a total of thirty compounds were found with the VIP value over 1.0. The entire VIP values are ranked in Table S3.

Response of Electronic Nose (E-Nose) Sensors to Volatile Organic Compounds (VOC) on Different Grades of Jasmine Tea
The signals of ten sensors in response to VOC are presented in Figure 3. The Figure shows the signal response of S1 and S2 was far stronger than the rest of the sensors (S3-S10). Indicated by analysis of variance (ANOVA), except for S3, there were significant differences existing between the different grade samples. Looking at trends of correlation, it was found that the response signals of S1, S2, S6, and S10 were negatively correlated with the sample grade, while the signals of S4, S5, S7, S8, and S9 were positively correlated with the grade, which suggests that S1, S2, S6, and S10 could respond to a grade-negative VOC, while, S4, S5, S7, S8, and S9 could respond to a grade-positive VOC.

Hierarchical Clustering Analysis
According to different correlation trends between the jasmine tea grades and the response signal intensity, all ten sensors can be subdivided into three categories, which were a negative correlation, positive correlation, and irrelevance.
Shown in the Frame A (Figure 4), for S1 (sensitive to Ammonia and Amines), S2 (Hydrogen sulfide and sulfides), S6 (Methane, ethane and hydrocarbons), and S10 (Alkanes and flammable gases), there was an apparent negative correlation between their signal intensity to the jasmine tea grade. The higher the response value was, in other words, the lower it was in its grade. The types of volatile organic compounds (VOC) they were sensitive to coincided with components negatively related to the jasmine tea quality. Heatmap of stable signals of E-nose sensors for six grades of jasmine tea (*Note: S1-S10 represent the ten sensors of the E-nose; 1G1, 1G2, 1G3, 1G4, 1G5, and 1G6 represent the six repeats of grade 1 jasmine tea, as do the following grade samples; The blue frame (A) indicates sensors with a negative correlation to the grade; The red frame (B) indicates sensors with a positive correlation to the grade).
The second type, as indicated in Frame B (Figure 4), includes S4 (Alcohol and Organic Solvents), S5 (Volatile gases in food cooking), S7 (Flammable gases) and S8 (Volatile Organic Compounds) as their signal response intensity was positively correlated with the tea grading, which meant they reflected the content of volatiles positively related to the jasmine tea grade, so we could name them as positive VOC recognition sensors.
The rest of the sensors, S3 (hydrogen) and S9 (Hydroxide, gasoline, and kerosene), were irrelevant sensors for evaluating jasmine tea aroma, as there was no regularity in the signals appearing in response to the grade. This result was reasonable because the corresponding sensitive gas does not exist in jasmine tea at all. Therefore, both S3 and S9 should be ignored to reduce data noise.

Principal Component Analysis (PCA)
After removing signals from both the S3 and S9 sensors, the electronic nose (E-nose) data was subjected to PCA analysis, through which we could obtain an overview of sample similarity. Shown in Figure 5, PC1 and PC2 explain 59.9% and 33.1% of the total variance, respectively. It was intriguing to find that the 1G, 2G, 3G, and 4G samples were not distinguished completely, whereas there was a clear separation of 5G and 6G samples from the other grade samples. After comparing the difference in sensory evaluation criteria of these grade samples, we found that it was reasonable. Rather than a significant difference in the aroma intensity [6] (Table S1), the main difference for samples in area I (1G, 2G, 3G, and 4G) were certain specific characteristics, such as the freshness and durability of the aroma. Therefore, it indicates that areas I (1G, 2G, 3G, and 4G), II (5G), and III (6G) were mainly reflecting aroma concentration. It also suggests that the E-nose could be good at recognizing aroma concentration but may not good at identifying specific unique aroma characteristics of high-grade jasmine tea. The following two reasons may attribute to this conclusion. First, the strength of volatile components that have a pivotal contribution to freshness and persistence was deficient and could not respond well to these sensors. Second, the formation of freshness and durability were not determined by some specific volatile substances, but by the combination of some elements within a particular range of proportion.

Conclusions
A group of authoritative jasmine tea grade samples, which were prepared following Chinese National Standard requirements, were subjected to research. Both Automatic thermal desorption-gas-chromatography-mass spectrometry (ATD-GC-MS) and electronic nose (E-nose) were applied for discrimination and were compared systematically.
Consequently, a total of sixty-three volatile compounds were tentatively identified by ATD-GC-MS. Through both partial least square-discriminant analysis (PLS-DA) and hierarchical cluster analysis (HCA), a satisfactory discriminant result was achieved. Twelve of these compounds, including four alcohols, five aldehydes, two hydrocarbons, and one ketone, were found to be negatively correlated to the jasmine tea grade. It is worth noting that most of the main aroma components of jasmine tea, such as linalool, (Z)-3-Hexenyl benzoate, methyl salicylate, (Z)-3-Hexenyl acetate, α-Farnesene and indole, have no linear relationship between their contents to the tea grade, but are obviously abundant in the high grade.
Regarding the electronic nose, the signal intensities of S1 (sensitive to Ammonia and Amines), S2 (Hydrogen sulfide and sulfides), S6 (Methane, ethane and hydrocarbons), and S10 (Alkanes and flammable gases) were negatively correlated to the tea grades. While, S4 (Alcohol and Organic Solvents), S5 (Volatile gases in food cooking), S7 (Flammable gases) and S8 (Volatile Organic Compounds), were positively correlated to the tea grades. It was interesting to find that the E-nose was better at detecting aroma concentrations rather than recognizing unique aroma characteristics.
Supplementary Materials: The following are available online at http://www.mdpi.com/1420-3049/25/2/380/s1; Figure S1: The processing steps of standard Chinese jasmine tea grade samplestitle, Figure S2: Total ion chromatogram diagram of VOCs with different grades jasmine tea, Table S1: Description of aroma characteristic in each standard jasmine tea sample, Table S2: Gas sensors array and corresponding volatile components, Table S3: VIP score of identified VOCs, Table S4. VOCs of grade jasmine tea.