Metabolomics Profiling of Roasted Stem Tea Using Gas Chromatography-Mass Spectrometry and a Sensory Test

Abstract

Roasted tea is manufactured by roasting green tea at a high temperature and has characteristic flavors (roasted, savory, or sweet) that complement Japanese-style food. Here, we evaluated the sensory quality for six kinds of roasted stem tea products using a sensory test, the rank-rating method, of 52 adolescent panelists. We then elucidated and analyzed the chemical components in these products by gas chromatography-mass spectrometry and principal component analysis. We found that the preferred products were located near the center of the principal component analysis score plot, indicating that the position in the plot was associated with the sensory score of roasted stem tea. We then used these findings to build a predictive model of sensory scores using orthogonal partial least squares regression analysis. This showed that the predicted sensory score was higher when certain pyrazines were present and lower when aldehydes and sugars were present. This is the first study to demonstrate a relationship between the chemical components of roasted stem tea and its sensory score based on the palatability of adolescents for it using metabolomics profiling.

Introduction

Tea is one of the most popular beverages in the world. Global tea production has exceeded 5 million tonnes in 2014, and the trend is continuously increasing with an annual growth rate of 3.7% for black tea and 9.1% for green tea. Production levels are expected to reach global totals of 4.29 million tonnes of black tea and 3.74 million tonnes of green tea by 2024 (i). Green tea is mainly produced in Asian countries such as China, Vietnam and Japan by steaming and drying raw tea leaves immediately after harvest. This steam treatment results in the inactivation of endogenous enzymes so that most of the leaf polyphenols remain unoxidized, giving green tea its antioxidant properties (Cabrera et al., 2006).

Roasted tea, which is obtained by roasting green tea, is very popular among Japanese consumers. People of all ages enjoy roasted tea for its characteristic flavor. The tea is roasted at a high temperature that is similar to that used to roast coffee, which generates the flavor, and diminishes its astringency. Roasted tea is usually consumed after a meal because its weak astringency complements Japanese-style food and its comforting flavor makes people relax. Like other types of green tea, roasted tea has also been reported to have high antioxidant activity (Satoh et al., 2005), despite the roasting process reducing its polyphenol contents (Nakagawa, 1967).

In Japan, roasted tea is typically made using lower grades of green tea harvested as the third- or fourth-flush teas between summer and autumn. Roasted tea is typically preferred as a reasonably priced tea. Notably, however, roasted stem tea is the preferred tea in Kanazawa, the capital city of Ishikawa Prefecture in Japan, which is renowned for being a consumption area of roasted tea. Roasted stem tea is obtained by roasting tea stems, a by-product of the manufacturing process of a particular green tea called Sencha. Tea stems are a suitable ingredient for roasted tea because they generate much more flavor during the roasting process than the leaves (Sasaki et al., 2017). Some roasted stem tea products are produced from the stems of the first-flush tea, and these are known for being used in high-grade tea products.

Roasted tea contains over 100 volatile components, such as pyrazines, pyrroles, and furans (Yamanishi et al., 1973; Kawakami and Yamanishi, 1999), with pyrazines in particular being one of the major odorants that affect its flavor (Mizukami et al., 2008; Sasaki et al., 2017). However, there have been few reports on flavor and taste components in roasted tea, and the relationship between chemical components and sensory score of roasted tea has never been reported. It is necessary to elucidate which compounds are important in governing the flavor and taste of roasted tea for its further improvement.

Metabolomics is a useful technique for identifying which components significantly affect food quality (Cevallos-Cevallos et al., 2009; Putri et al., 2013). Metabolomics-based models for the prediction of green tea quality have been developed using gas chromatography-mass spectrometry (GC/ MS), liquid chromatography-mass spectrometry (LC/MS), and capillary electrophoresis-mass spectrometry (CE/MS), which have shown that quinic acid, theanine, and sugars are important components (Pongsuwan et al., 2007, 2008). Application of a metabolomics approach to roasted tea production is expected to reveal the influence of chemical components on quality.

The main objective of this study was to elucidate the relationship between the sensory score for roasted stem tea by adolescents and chemical components of roasted stem tea using metabolomics profiling, and to develop a predictive method for obtaining the sensory score based on chemical components. We evaluated sensory score for adolescent panelists using the rank-rating method, and analyzed the volatile and nonvolatile components using GC/MS with stir bar sorption extraction (SBSE) and trimethylsilylation (TMS). Next, we tried to obtain a comprehensive understanding of the chemical components of roasted stem tea using principal component analysis (PCA), especially through an understanding of the influence of roasting degree and grade of roasted stem tea product on component profile. We then examined whether the sensory score could be predicted by using a prediction model that constructed using orthogonal projections to latent structures (OPLS) regression analysis.

Materials and Methods

Materials    Tea products: We used six kinds of roasted stem tea products, all of which were obtained from a local market in Kanazawa, Ishikawa, Japan. The price range of the products was 150–1350 yen/100 g (150 yen/100 g: 1 product, 400 yen/100 g: 1 product, 1000–1350 yen/100 g: 4 products). Roasting categories of the products were light, medium and dark, which were classified based on sensory tests for taste, flavor and color by five specialists of tea manufactures. Chemicals: For the analysis of volatile compounds, we purchased cyclohexanol and sodium chloride from Wako Pure Chemical Industries Ltd. (Tokyo, Japan) and N-alkanes (n = 6–25) from Tokyo Chemical Industry Co., Ltd. (Tokyo, Japan). For the analysis of nonvolatile compounds, we purchased methanol, chloroform, ribitol, and pyridine from Wako Pure Chemical Industries Ltd.; methoxyamine hydrochloride from Sigma-Aldrich Co. LLC. (WO, USA); and N-methyl-N-(trimethylsilyl) trifluoroacetamide from GL Science, Inc. (Tokyo, Japan). For the identification of compounds detected by GC/MS, we purchased volatile compounds 1, 2, 5–9, 12, 13, 16, 19–24, 29–33, 35–37, and 40, and nonvolatile compound 21 from Tokyo Chemical Industry Co., Ltd.; volatile compound 36 and nonvolatile compounds 1–16 from Wako Pure Chemical Industries Ltd.; nonvolatile compounds 17–20 from Nagara Science Co., Ltd. (Gifu, Japan); and volatile compounds 38 and 39 from Sigma-Aldrich Co. LLC (see Tables 2 and 3 later in the report for a list of these compounds).

Table 2. Characteristics of the 40 identified volatile compounds. (Note: An additional 43 volatile compounds remained unidentified.)

No.	RTa	RIb	Compound	IDc	Type
1	18.3	1076	Hexanal	MS(a,d), RI(a,l1)	Aldehyde
2	22.2	1178	1-Ethylpyrrole	MS(a,d), RI(a,l2,3)	Pyrrole
3	22.4	1182	2-Heptanone	MS(d), RI(l2)
4	23.1	1200	2,4,5-Trimethyloxazole	MS(d), RI(l4)
5	25.4	1269	2-Methylpyrazine	MS(a,d), RI(a,l4)	Pyrazine
6	27.3	1329	2,5-Dimethylpyrazine	MS(a,d), RI(a,l2,3)	Pyrazine
7	27.5	1335	2,6-Dimethylpyrazine	MS(a,d), RI(a,l2)	Pyrazine
8	27.6	1339	2-Ethylpyrazine	MS(a,d), RI(a,l3,4)	Pyrazine
9	27.6	1338	6-Methyl-5-hepten-2-one	MS(a,d), RI(a,l5)
10	29.2	1392	2-Ethyl-6-methylpyrazine	MS(d), RI(l3,6)	Pyrazine
11	29.4	1399	2-Ethyl-5-methylpyrazine	MS(d), RI(l6)	Pyrazine
12	29.8	1412	2-Ethyl-3-methylpyrazine	MS(a,d), RI(a,l6)	Pyrazine
13	29.9	1415	2,3,5-Trimethylpyrazine	MS(a,d), RI(a,l3,7)	Pyrazine
14	30.8	1450	(E)-linalool oxide (furanoid)	MS(d), RI(l3,8)	Terpene
15	30.9	1455	2-Ethyl-3,6-dimethylpyrazine	MS(d), RI(l3,8)	Pyrazine
16	31.2	1462	Furfural	MS(a,d), RI(a,l8)	Aldehyde
17	31.3	1473	(E,Z)-2,4-heptadienal	MS(d), RI(l9)	Aldehyde
18	31.4	1472	2-Ethyl-3,5-dimethylpyrazine	MS(d), RI(l3,8)	Pyrazine
19	31.8	1487	2-Ethyl-1-hexanol	MS(a,d), RI(a)
20	32.2	1500	(E,E)-2,4-heptadienal	MS(a,d), RI(a,l9)	Aldehyde
21	32.3	1504	2,3-Diethyl-5-methylpyrazine	MS(a,d), RI(a,l3,8)	Pyrazine
22	33.0	1531	Benzaldehyde	MS(a,d), RI(a,l1)	Aldehyde
23	33.3	1547	Linalool	MS(a,d), RI(a,l3,6)	Terpene
24	34.2	1576	5-Methyl furfural	MS(a,d), RI(a,l3,6)	Aldehyde
25	34.9	1607	Hotrienol	MS(d), RI(l7)	Terpene
26	35.2	1622	1-Ethyl-2-formylpyrrole	MS(d), RI(l10)	Pyrrole
27	39.2	1790	Methyl salicylate	MS(a,d), RI(a,l3,8)
28	40.0	1833	1-Furfurylpyrrole	MS(a,d), RI(a,l3,6)	Pyrrole
29	40.3	1851	Geraniol	MS(a,d), RI(a,l3,11)	Terpene
30	42.3	1941	Benzyl cyanide	MS(a,d), RI(a,l5)
31	42.7	1962	β-Ionone	MS(a,d), RI(a,l5)	Terpene
32	42.9	1971	(Z)-jasmone	MS(a,d), RI(a,l3,11)
33	43.1	1982	2-Acetylpyrrole	MS(a,d), RI(a,l10)	Pyrrole
34	43.9	2019	5,6-Epoxy-β-ionone	MS(d), RI(l10)	Terpene
35	46.6	2162	Nonanoic acid	MS(a,d), RI(a,l10)
36	47.4	2206	2-Methoxy-4-vinylphenol	MS(a,d), RI(a,l6,11)
37	49.1	2297	Jasmine lactone	MS(a,d), RI(a,l3,11)
38	49.9	2336	Geranic acid	MS(a,d), RI(a,l3)	Terpene
39	51.1	2394	4-Vinylphenol	MS(a,d), RI(a)
40	52.6	2459	Indole	MS(a,d), RI(a,l3,11)

^a  RT: retention time.

^b  RI: retention index on a DB-Wax column (60 m × 0.25 mm i.d., coated with a film thickness of 0.25 µm) observed for gas chromatography-mass spectrometry (GC/MS).

^c  Method of identification: MS(a), identification by comparison with mass spectra of authentic reference compounds; MS(d), identification by comparison with mass spectra stored in the National Institute of Standards and Technology (NIST) and Wiley databases; RI(a), identification by comparison with the RIs of authentic reference compounds; RI(l), identification by comparison with published RIs in the literature (¹, Mattheis et al., 2005; ², Shimoda et al., 1996a; ³, Sasaki et al., 2017; ⁴, Shimoda et al., 1996b; ⁵, Tatsuka et al., 1990; ⁶, López-Galilea et al., 2006; ⁷, Ito et al., 2002; ⁸, Mizukami et al., 2008; ⁹, Buttery et al., 2001; ¹⁰, Shimoda et al., 1995; ¹¹, Kumazawa and Masuda, 2002).

Table 3. Characteristics of the 21 identified nonvolatile compounds. (Note: An additional 26 nonvolatile compounds remained unidentified.)

No.	RT	Compound	Type
1	5.6	Oxialic acid	Organic acid
2	7.0	Phosphoric acid
3	7.8	Serine	Amino acid
4	9.0	Malic acid	Organic acid
5	9.3	Aspartic acid	Amino acid
6	9.5	Pyroglutamic acid	Amino acid
7	10.1	Glutamic acid	Amino acid
8	11.2	Theanine	Amino acid
9	11.4	Shikimic acid	Organic acid
10	11.5	Citric acid	Organic acid
11	11.7	Quinic acid	Organic acid
12	12.0	Glucose	Sugar
13	12.4	Caffeine
14	12.5	Gallic acid	Organic acid
15	13.3	Inositol	Sugar
16	16.2	Sucrose	Sugar
17	17.5	Epicatechin	Catechin
18	17.6	Catechin	Catechin
19	17.7	Gallocatechin	Catechin
20	17.8	Epigallocatechin	Catechin
21	17.8	Galactinol	Sugar

Nonvolatile compounds were identified by comparison with the mass spectra and retention times (RTs) of authentic compounds.

Sensory evaluation    Fifty-two panelists (18 men and 34 women, aged 20–23 years) were recruited from Ishikawa Prefectural University, Nonoichi, Japan. The sensory quality of roasted stem tea products for non-trained panelists was evaluated by rank-rating method, which was developed for inexperienced sensory panelists by O'Mahony et al. (Kim and O'Mahony, 1998). The sensory test was performed in a sensory room that was divided into four panelist areas. Dried tea leaves of each product (6–10 g, based on an extraction recipe recommended by the respective manufacturers) were extracted with 300 mL of boiling water in a teapot for 30 seconds. The boiled water was produced from bottled water (Oishii Mizu Rokko: Asahi Soft Drinks Co. Ltd., Tokyo, Japan), the hardness of which is 32 mg/L (soft water level). The extract was then filtered through a stainless tea strainer into another teapot and 50 mL of the filtrate was poured from the pot into a 90-mL paper cup that was labeled with a three-digit random number.

Each of the hot filtrates was served to the panelists at the same time, and they ranked their quality. Rank-rating was carried out using a numerical rating scale that was divided into 11 parts numbered 1–11, and was labeled “more” at one end and “less” at the other, with a higher score indicating a greater preference from the panelist. The evaluation used the same scaling sheet for all of the products. Panelists were able to re-taste each of the products if they so desired and could alter their rating score on the sheet. Differences between product scores were then analyzed using Wilcoxon signed rank test.

Volatile compound analysis by GC/MS with the SBSE method    Volatile compounds in the tea brews were analyzed using the SBSE method, as described previously (Sasaki et al., 2017). Tea brew (10 mL) was added to cyclohexanol as an internal standard to give a final cyclohexanol concentration of 1 mg/L. NaCl (3 g) was then added to the tea brew and dissolved. A 15-mm stir bar with a 500-µm polydimethylsiloxane layer was added to the tea brew and equilibrated for 1 h at room temperature with constant stirring. The stir bar was then removed from the tea brew and dried on filter paper.

For the GC/MS analyses, the stir bar was inserted into a thermal desorption unit (TDU; Gerstel Inc., Mülheim an der Ruhr, Germany) that was connected to a cold trap injector (CIS; Gerstel Inc.) and a GC system. The extracted odorants were desorbed from the stir bar at 230°C using the thermal desorption system, cryofocused at −120°C using liquid N₂, volatilized at 230°C in the CIS, and injected into a GC system (Agilent 7890A; Agilent Technologies Inc., Palo Alto, CA, USA) equipped with a mass spectrometer (Agilent 5975C; Agilent Technologies Inc.). The column was a 60 m × 0.25 mm internal diameter (i.d.) DB-Wax fused-silica capillary column (J&W Scientific Inc., Folsom, CA, USA) with a film thickness (df) of 0.25 µm. The oven temperature was programmed to increase from 40°C (10-min hold) to 230°C (12-min hold) at a rate of 5°C/min. The split ratio was 1:10. Helium was used as the carrier gas at a linear flow rate of 1.2 mL/min. The mass spectra were obtained by electron-impact ionization under an ionization voltage of 70 eV and an ion source temperature of 150°C. Analysis was carried out using the SCAN mode. The retention index (RI) was calculated in accordance with the method of Kováts (1958) using N-alkanes (n = 6–25). Each product was analyzed in triplicate.

Nonvolatile compound analysis by GC-MS with the silylation method    Nonvolatile compounds were analyzed according to methods described in previous research (Pongsuwan et al., 2007). Roasted tea was crushed with a ball mill (µT-01; TAITEC Corp., Saitama, Japan) containing 3.0-mm zirconium oxide grinding beads at 4,600 rpm for 1 min. The resulting powder (15 mg) was then placed in a 1.5-mL micro tube, and 1 mL of a methanol, distilled water, and chloroform mixture in a ratio of 2.5:1:1 (v/v/v) was added and mixed with the sample. The mixture was shaken with 60 µL ribitol (diluted to 0.2 mg/mL in deionized water) as an internal standard for 5 min and was centrifuged at 16,000 × g for 3 min. The supernatant (400 µL) was then transferred to another 1.5-mL micro tube. The extract was dried by flowing dry N₂ gas over it for 1 h and then placing it in an incubator at 60°C overnight.

Oxidation and trimethylsilylation were used for derivatization. First, 50 µL of methoxyamine hydrochloride adjusted to 20 mg/mL in pyridine was added to the dried sample. The mixture was reacted at 30°C for 90 min, and 100 µL of N-methyl-N-(trimethylsilyl) trifluoroacetamide was then added and the mixture was reacted again at 37°C for 30 min.

GC/MS analysis was performed on a gas chromatography instrument (Agilent 6890 N; Agilent Technologies Inc.) equipped with a mass selective detector (Agilent 5975; Agilent Technologies Inc.). The column was a 5% phenyl polysilphenylenesiloxane capillary column (SGE ENV-5MS; Kanto Chemical Co. Inc., Tokyo, Japan) with a 30 m × 0.25 mm i.d. and 1.0 µm df. The derivatized sample (1 µL) was injected in split mode (25:1, v/v) with an injection temperature of 250°C. Helium was used as the carrier gas at a linear flow rate of 1.1 mL/min. The oven temperature was programmed to increase from 80°C (2-min hold) to 330°C (6-min hold) at a rate of 15°C/min The mass spectra were obtained by electron-impact ionization under an ionization voltage of 70 eV and an ion source temperature of 250°C. Analysis was carried out using the SCAN mode. Each product was analyzed in triplicate.

Identification    Volatile compounds were identified by comparison with reference mass spectra stored in the Wiley and National Institute of Standards and Technology (NIST) databases, the mass spectra and RIs of authentic compounds, or the RIs reported in published papers. Nonvolatile compounds were identified by comparison with the mass spectra and retention times (RTs) of authentic compounds.

Results and Discussion

Sensory evaluation    Table 1 shows the results of the sensory evaluation of six roasted stem tea products (nos. 1–6) by 52 adolescent panelists using the rank-rating method and their respective sensory scores. On the basis of these scores, the products were divided into three groups: high (products no. 1 and no. 2), middle (products no. 3 and no. 4), and low (products no. 5 and no. 6). Results from a Wilcoxon signed rank test (P < 0.05) revealed some significant differences. For example, product no.1 was significantly different from products no. 5 and no. 6 in the low score group, and product no.6 was significantly different from products no. 1 and no. 2 in the high score group.

Table 1. Results of the sensory evaluation.

Product rank	Score	Roasting Characteristics	Gradea
1	7.1 ± 2.4A	Medium	High
2	7.1 ± 2.6AB	Dark	Medium
3	6.4 ± 2.9ABC	Light	High
4	6.3 ± 3.2ABC	Light	High
5	5.4 ± 2.9BC	Medium	High
6	5.3 ± 3.3C	Dark	Low

^a  Grade was categorized on the basis of product price: High, 1000–1350 yen/100 g, Medium: 400 yen/100 g, Low: 150 yen/100 g

All values are means ± standard errors. Different upper-case letters indicate significant differences at P < 0.05 (Wilcoxon signed rank test). The sensory evaluation was carried out by 52 panelists, with a high score meaning that the product was preferred by the panelists.

The high and low score groups included both the medium-and dark-roasted stem tea, whereas the middle score group contained only the light-roasted stem tea. Light-roasted stem tea is often used as a gift, to serve to guests, or to drink on special occasions. By contrast, the medium- and dark-products are consumed in daily life, such as after a meal, with Japanese sweets, and for a relaxing break. Therefore, it is possible that the panelists did not select the light-products as the preferred teas because they were less familiar with these products. Product no. 2 was ranked at a high level, even though it was a medium grade product. Additionally, product no. 5 was ranked at a low level, despite it being a high grade product. These results show that product price did not correspond with preference level. In contrast to this trend, product no. 6, a low grade product, was ranked at a low level. As generally expected for food quality, panelists did not prefer low price products.

Component profiling    GC/MS with the SBSE and silylation methods detected a total of 83 volatile and 47 nonvolatile compounds across the six tea products, 61 of which could be identified. These included 11 pyrazines, 4 pyrroles, 7 terpenes, 6 aldehydes, 6 organic acids, 5 amino acids, 4 sugars, and 4 catechins (Tables 2, 3). The combined data for these 130 compounds were analyzed by PCA, the results of which are shown in Fig. 1. In the score plot (Fig. 1A), the first two principal components, PC1 and PC2, explained 40.1% and 19.4% of the variance, respectively. By contrast, PC3 and PC4 explained 11.2% and 8.0%, respectively, with the replicate data points for each product being dispersed along these axes, and so they are not shown on the figure. Products no. 1–4, high and medium sensory score products, were located near the center of the plot (Fig. 1A). Panelists preferred those products that were placed in the center of the plot. Product no. 5 was located in the high negative PC2 area, and product no. 6 was located in the high positive PC1 area. Low sensory score products were located further away from the center of the plot, indicating that the components of the products differed between high and medium sensory score products. The position on the plot was associated with the sensory score of roasted stem tea.

Fig. 1.

Principal component analysis of the compounds found in six products of roasted stem tea: (A) score plot and (B) loading plot.

Products that received high, middle, and low scores in the sensory evaluation are represented by black, gray, and open symbols, respectively. Compounds that are classified as pyrazines, pyrroles, terpenes, aldehydes, organic acids and others are represented by gray circles, open circles, gray squares, open diamonds, open triangles and black circles, respectively. The degree of roasting (light, medium, or dark) (A) and the names of the primary volatile and nonvolatile compounds (B) are indicated alongside the symbols. The mean values obtained from three independent samples per product are shown in A.

The light-roasted (nos. 3 and 4) and medium-roasted (nos. 1 and 5) stem tea products were located in the negative PC1 area, whereas the dark-roasted products (nos. 2 and 6) were located in the positive PC1 area, indicating that the components that were associated with the PC1 axis were related to roasting type. The roasting process generates many odorants of roasted stem tea through the Maillard reaction between amino acids and sugars, giving it a roasted odor (Mizukami et al., 2008; Sasaki et al., 2017). In the loading plot (Fig. 1B), 10 pyrazines were located in the negative PC1 area, some of which had highly negative scores. This indicates that the light- and medium-roasted products contained more pyrazines than the dark-roasted products, which is consistent with the fact that light-roasted stem tea usually has a strongly roasted flavor. Pyrroles are also generated through the Maillard reaction and were also located in the negative PC1 area of the loading plot.

The roasting process decomposes many components in green tea, but more lightly roasted stem tea will retain the major components, such as catechins, caffeine, amino acids, and organic acids. Lightly roasted stem tea is generally made using higher grades of green tea harvested early as the first tea, whereas dark roasted stem tea is typically made using the lower grades harvested as the third- or fourth-flush teas between summer and autumn. It also influences the amount of components in roasted stem tea. With regard to catechins, epigallocatechin (EGC), epicatechin (EC), gallocatechin (GC), and catechin (C) were located in the negative PC1 area of the loading plot, indicating that lightly roasted stem tea contained larger amounts of catechins. It has been reported that the roasting process decomposes catechins in green tea (Mizukami et al., 2008). However, ingredients for light-roasted products are expected to contain smaller amounts of catechins than ingredients for dark roasted products, because the first green tea contains smaller amounts of catechins than the third green tea (Anan et al., 1974). Therefore, mainly due to the influence of the roasting process, lightly roasted stem tea contains larger amounts of catechins. Additionally, the roasting process isomerizes EGC, EC, epigallocatechin gallate (EGCg), and epicatechin gallate (ECg), which are usually found in green tea (Yamamoto et al., 1997), to produce GC, C, gallocatechin gallate (GCg), and catechin gallate (Cg) (Mizukami et al., 2008) (although note that EGCg, ECg, GCg, and Cg were not detected in this analysis). Consequently, the PC1 loading score of GC and C was higher than that of EGC and EC in the negative PC1 direction.

Most of the amino acids (theanine, asparagine, serine, and pyroglutamic acid) and organic acids (citric acid, malic acid, etc.) were located in the negative PC1 area of the loading plot, indicating that lightly roasted stem tea contained larger amounts of amino acids and organic acids. This makes sense because the roasting process decomposes amino acids. The ingredients for light-roasted products are also expected to contain larger amounts of amino acids than ingredients for dark roasted products, because the first green tea contained large amounts of amino acids (Anan et al., 1974). These factors are thought to be the reasons for lightly roasted stem tea containing larger amount of amino acids. With regard to organic acids, the amounts of organic acids are decreased when harvest is late, however the difference is only slight (Horie et al., 2002) The roasting process that decomposes organic acids is thought to mainly influence the amount of organic acids.

Product no. 6 had a high positive score along the PC1 axis (Fig. 1A), and aldehydes (2,4-heptadienal, hexanal furfural, 5-methylfurfural) and sugars (glucose, galactinol, inositol) were also found to have a high positive loading score along the PC1 axis (Fig. 1B). Product no. 6 contained larger amounts of (E,Z)-2,4-heptadienal and (E,E)-2,4-heptadienal than the other products, which are known to be off-flavor odorants (Hara et al., 1974; Hashizume et al., 2007; Fujita et al., 2010), indicating that heptadienal has a bad influence on the quality of this product. Heptadienal is generated from the linolenic acid in tea through oxidation (Hara and Kubota, 1982) or cleavage by hydroperoxide lyase (Yang et al., 2013). Furthermore, product no. 6 contained more sugars such as glucose than the other products, even though sugars are typically decomposed by the dark roasting process. These findings indicate that product no. 6 may have been produced from autumn-harvested green tea, as this contains more sugars and linolenic acid than spring- and early summer-harvested teas (Yang et al., 2013). Furfural and 5-methylfurfural possess an almond odor and are the major odorants in many foods and beverages, such as bread, baked potato, and whiskey. It has been reported that 5-methylfurfural in roasted tea can be detected by gas chromatography-olfactometry (GC-O) (Mizukami et al., 2008). Therefore, 5-methylfurfural may influence the sensory score of roasted stem tea. Because furfural and 5-methylfurfural are generated through the dehydration of monosaccharides, which are more present in higher levels in autumn-harvested green tea, it appears that the dark-roasted product no. 6 contained more furfural and 5-methylfurfural than the other roasted stem tea products.

With regard to terpenes, geraniol, linalool, and hotrienol were located in the negative PC1 area on the loading plot, whereas β-ionone, 5,6-epoxy-β-ionone, and (E)-linalool oxide were located in the positive PC1 area. It is expected that these compounds give the roasted stem tea products a floral flavor because they exhibit floral odors. (E)-linalool oxide is generated through the oxidation of linalool and so exhibits a positive PC1 score.

Product no. 5 had a highly negative PC2 score in the PCA plot, and pyrroles, geranic acid, and methyl salicylate also exhibited negative PC2 loading scores. Some kinds of pyrroles have a roasted, nutty, and sweet odor in tea (Hattori et al. 2005; Sasaki et al., 2017). Geranic acid exhibits a floral odor, and methyl salicylate has a medicinal odor. Moreover, product no. 5 contained lower levels of amino acids and catechins than product nos. 1–4, which exhibited high scores in the positive PC2 direction. These were characteristic compounds contained in product no. 5 which received a low sensory score. However, the influence of these compounds on quality was unclear.

Development of a ranking predictive model by partial least squares    To investigate the relationship between the sensory score and component profile of each product, we developed a ranking predictive model using OPLS. The OPLS predictive model was constructed using the volatile and nonvolatile profiles (Fig. 2) and the sensory scores (Table 4), using product no. 4 as a training sample. There was a strong positive linear correlation between the observed and predicted scores for each product (r = 0.9992).

Fig. 2.

Relationship between the measured and predicted sensory qualities for roasted stem tea measured using an orthogonal partial least squares (OPLS) model.

Products that received high, middle, and low scores in the sensory evaluation are represented by black, gray, and open symbols, respectively. Product no. 4 was used as a training sample and the other products were used as calibration samples. The mean values ± standard errors obtained from three independent samples per product are shown.

Table 4. The key compounds that were found to have variable importance in projection (VIP) values >1.2 and their coefficients in the orthogonal partial least squares (OPLS) regression model.

VIP	Coefficient	Compound	Volatility
Positive direction
1.45	0.044	2,6-Dimethylpyrazine	Volatile
1.38	0.027	Hotrienol	Volatile
1.38	0.020	Unknown	Volatile
1.29	0.047	2-Methyl pyrazine	Volatile
1.28	0.019	2,5-Dimethylpyrazine	Volatile
1.27	0.018	Unknown	Volatile
1.26	0.027	Caffeine	Nonvolatile
1.25	0.019	2-Ethyl-3-methylpyrazine	Volatile
1.25	0.025	2-Ethyl-6-methylpyrazine	Volatile
Negative direction
1.58	−0.045	Unknown	Volatile
1.48	−0.038	Inositol	Nonvolatile
1.46	−0.032	Methyl salicylate	Volatile
1.45	−0.029	5-Methyl furfural	Volatile
1.44	−0.036	Furfural	Volatile
1.38	−0.024	Unknown	Nonvolatile
1.29	−0.025	Unknown	Nonvolatile
1.29	−0.029	Glucose	Nonvolatile
1.28	−0.016	Unknown	Nonvolatile
1.26	−0.027	(E,Z)-2,4-heptadienal	Volatile
1.22	−0.032	Unknown	Volatile
1.21	−0.029	Unknown	Nonvolatile

The key compounds that had a variable importance in projection (VIP) value of >1.2 are listed in Table 4. Positive or negative values were determined on the basis of their coefficients. Pyrazines, hotrienol, and caffeine were found to have high VIP values and positive coefficient scores. Pyrazines, which was strongly detected by GC-O, are present in expensive roasted tea products (Mizukami et al., 2008); hotrienol is found in high-quality oolong tea and has a floral odor (Kawakami et al., 1995; Cho et al., 2007); and caffeine is also present at high levels in award-winning green teas at the Japanese Tea Competition (Horie et al., 1993). Therefore, the presence of these compounds may result in high sensory scores for some roasted stem tea products.

Some volatile compounds [(E,Z)-2,4-heptadienal, 5-methylfurfural, methyl salicylate and furfural], and nonvolatile compounds (glucose and inositol) were also found to have high VIP values but negative coefficients (Table 4). As mentioned previously, (E,Z)-2,4-heptadienal is known to be an off-flavor odorant (Hara et al., 1994; Hashizume et al., 2007; Fujita et al., 2010). Furthermore, although 5-methylfurfural, furfural, and methyl salicylate are often used to add flavor to foods and beverages, they would also cause an off-flavor at high concentrations. Sugars are expected to have little effect on the taste of tea but may indicate that autumn-harvested tea was used to produce the roasted stem tea.

Conclusion

In this study, we conducted a sensory evaluation on adolescents by using the rank-rating method and used metabolomics profiling with GC/MS to gain a comprehensive understanding of the sensory score for and chemical components of six different roasted stem tea products. The preferred products in the sensory evaluation were located near the center of the PCA score plot, indicating that the position in the plot was associated with the sensory score of roasted stem tea. The PC1 axis in the PCA score plot was related to roasting type, with components such as pyrazines and catechins being located here. The predictive model for sensory scores generated by OPLS regression analysis demonstrated that there was a strong positive linear relationship between the observed and predicted scores, and verification of the VIP values in the model identified some compounds that were related to the sensory score. Thus, this study successfully revealed the relationship between sensory score for the younger generation and the chemical components of roasted stem tea.

References

•  Anan,  T. and  Nakagawa,  M. (1974). Effect of light on chemical constituents in the tea leaves. J. Agric. Chem. Soc. Japan (Nippon Nōgeikagaku Kaishi), 48, 91-96. (In Japanese).
•  Buttery,  R. G.,  Takeoka,  G. R.,  Naim,  M.,  Rabinowith,  H., and  Nam,  Y. (2001). Analysis of furaneol in tomato using dynamic headspace sampling with sodium sulfate. J. Agric. Food Chem., 49, 4349-4351.
•  Cabrera,  C.,  Artacho,  R., and  Giménez,  R. (2006). Beneficial effects of green tea--a review. J. Am. Coll. Nutr., 25, 79-99.
•  Cevallos-Cevallos,  J. M.,  Reyes-De-Corcuera,  J. I.,  Etxeberria,  E.,  Danyluk,  M. D. and  Rodrick,  G. E. (2009). Metabolomic analysis in food science: a review. Trends Food Sci. Technol., 20, 557-566.
•  Cho,  J. Y.,  Mizutani,  M.,  Shimizu,  B.,  Kinoshita,  T.,  Ogura,  M.,  Tokoro,  K.,  Lin,  M.L., and  Sakata,  K. (2007). Chemical profiling and gene expression profiling during the manufacturing process of Taiwan oolong tea “Oriental Beauty”. Biosci. Biotechnol. Biochem., 71, 1476-1486.
•  Fujita,  A.,  Isogai,  A.,  Endo,  M.,  Utsunomiya,  H.,  Nakano,  S., and  Iwata,  H. (2010). Effects of sulfur dioxide on formation of fishy off-odor and undesirable taste in wine consumed with seafood. J. Agric. Food Chem., 58, 4414-4420.
•  Hara,  T. and  Kubota,  E. (1974). Changes in aroma components during heating of green tea. J. Jpn. Soc. Food Sci. Technol. (Nippon Syokuhin Kogyo Gakkaiish), 21, 495-498 (In Japanese).
•  Hara,  T. and  Kubota,  E. (1982). Changes in aroma components of green tea after storage. J. Agric. Chem. Soc. Japan (Nippon Nōgeikagaku Kaish)., 56, 625-630 (In Japanese).
•  Hashizume,  M.,  Gordon,  M. H., and  Mottram,  D. S. (2007). Light-induced off-flavor development in cloudy apple juice. J. Agric. Food Chem., 55, 9177-9182.
•  Hattori,  S.,  Takagaki,  H., and  Fujimori,  T. (2005). Identification of volatile compounds which enhance odor notes in Japanese green tea using the OASIS (original aroma simultaneously input to the sniffing port) method. Food Sci. Technol. Res., 11, 171-174.
•  Horie,  H.,  Mukai,  T., and  Goto,  T. (1993). Chemical composition of prized Japanese green tea in national tea competition 2. Total nitrogen, free amino acids, tannin, caffeine and ascorbic acid in ordinary Sen-cha, Fukamushi Sen-cha, Mushisei Tamaryoku-cha and Kamairisei Tamaryoku-cha. Tea Research Journal (Chagyo Kenkyu Hokoku), 77, 23-28.
•  Horie,  H.,  Ujihara,  T., and  Kohata,  T. (2002). Contents o organic acids in green teas and their contribution to tea quality. Tea Research Journal (Chagyo Kenkyu Hokoku), 93, 26-33.
•  Ito,  Y.,  Sugimoto,  A.,  Kakuda,  T., and  Kubota,  K. (2002). Identification of potent odorants in Chinese jasmine green tea scented with flowers of Jasminum sambac. J. Agric. Food Chem., 50, 4878-4884.
•  Kawakami,  M.,  Ganguly,  S. N.,  Banerjee,  J., and  Kobayashi,  A. (1995). Aroma composition of oolong tea and black tea by brewed extraction method and characterizing compounds of Darjeeling tea aroma J. Agric. Food Chem., 43, 200-207.
•  Kawakami,  M. and  Yamanishi,  T. (1999). Formation of aroma components in roasted or pan-fired green tea by roasting or pan-firing treatment. J. Agric. Chem. Soc. Japan (Nippon Nōgeikagaku Kaishi), 73, 893-906 (In Japanese).
•  Kim,  K.O. and  O'Mahony,  M. (1998). A new approach to category scales of intensity. I: Traditional versus rank-rating. J. Sens. Stud., 13, 241-249.
•  Kováts  E. (1958). Gas-chromatographische charakterisierung organischer verbindungen. teil 1: Retentionsindices aliphatischer halogenide, alkohole, aldehyde und ketone. Helvetica Chimica Acta., 41, 1915-1932.
•  Kumazawa,  K. and  Masuda,  H. (2002). Identification of potent odorants in different green tea varieties using flavor dilution technique. J. Agric. Food Chem., 50, 5660-5663.
•  López-Galilea,  I.,  Fournier,  N.,  Cid,  C., and  Guichard,  E. (2006). Changes in headspace volatile concentrations of coffee brews caused by the roasting process and the brewing procedure. J. Agric. Food Chem., 54, 8560-8566.
•  Mattheis,  J. P.,  Fan,  X., and  Argenta  L. C. (2005). Interactive responses of gala apple fruit volatile production to controlled atmosphere storage and chemical inhibition of ethylene action. J. Agric. Food Chem., 53, 4510-4516.
•  Mizukami,  Y.,  Sawai,  Y., and  Yamaguchi,  Y. (2008). Changes in the concentrations of acrylamide, selected odorants, and catechins caused by roasting of green tea. J. Agric. Food Chem., 56, 2154-2159.
•  Nakagawa,  M. (1967). The nature and the origin of polyphenols in Hoji-cha (Roasted Green Tea). Agr. Biol. Chem., 31, 1283-1287.
•  Pongsuwan,  W.,  Fukusaki,  E.,  Bamba,  T.,  Yonetani,  T.,  Yamahara,  T., and  Kobayashi,  A. (2007). Prediction of Japanese green tea ranking by gas chromatography/mass spectrometry-based hydrophilic metabolite fingerprinting. J. Agric. Food Chem., 55, 231-236.
•  Pongsuwan,  W.,  Bamba,  T.,  Harada,  K.,  Yonetani,  T.,  Kobayashi,  A., and  Fukusaki,  E. (2008). High-throughput technique for comprehensive analysis of Japanese green tea quality assessment using ultra-performance liquid chromatography with time-of-flight mass spectrometry (UPLC/TOF MS). J. Agric. Food Chem., 56, 10705-10708.
•  Putri,  S. P.,  Nakayama,  Y,  Matsuda,  F.,  Uchikata,  T.,  Kobayashi,  S.,  Matsubara,  A., and  Fukusaki,  E. (2013). Current metabolomics: Practical applications. J. Biosci. Bioeng., 115, 579-589.
•  Sasaki,  T.,  Koshi,  E.,  Take,  H.,  Michihata,  T.,  Maruya,  M., and  Enomoto,  T. (2017). Characterisation of odorants in roasted stem tea using gas chromatography-mass spectrometry and gas chromatographyolfactometry analysis. Food Chemistry, 220, 177-183.
•  Satoh,  E.,  Tohyama,  N., and  Nishimura,  M. (2005). Comparison of the antioxidant activity of roasted tea with green, oolong, and black teas. Int. J. Food Sci. Nutr., 56, 551-559.
•  Shimoda,  M.,  Shigematsu,  H.,  Shiratsuchi,  H., and  Osajima,  Y. (1995). Comparison of volatile compounds among different grades of green tea and their relations to odor attributes. J. Agric. Food Chem., 43, 1621-1625.
•  Shimoda,  M.,  Shiratsuchi,  H.,  Nakada,  Y.,  Wu,  Y., and  Osajima,  Y. (1996a). Identification and sensory characterization of volatile flavor compounds in sesame seed oil. J. Agric. Food Chem., 44, 3909-3912.
•  Shimoda,  M.,  Rossana,  R. P., and  Osajima,  Y. (1996b). Headspace gas analysis of fish sauce. J. Agric. Food Chem., 44, 3601-3605.
•  Tatsuka,  K.,  Suekane,  S.,  Sakai,  Y., and  Sumitani,  H. (1990). Volatile constituents of kiwi fruit flowers: simultaneous distillation and extraction versus headspace sampling. J. Agric. Food Chem., 38, 2176-2180.
•  Yamamoto,  T.,  Juneja,  L. R.,  Chu,  S. C., and  Kim,  M. (1997). “Chemistry and Applications of Green Tea.” CRC Press, Boca Raton.
•  Yamanishi,  T.,  Shimoji,  S.,  Ukita,  M.,  Kawashima,  K., and  Nakatani,  Y. (1973). Aroma of roasted green tea (Hoji-cha). Agric. Biol. Chem., 37, 2147-2153.
•  Yang,  Z.,  Baldermann,  S., and  Watanabe,  N. (2013). Recent studies of the volatile compounds in tea. Food Res. Int., 53, 585-599.
• http://www.fao.org/fileadmin/templates/est/COMM_MARKETS_MONITORING/Tea/IGG22/16-Inf2-CurrentSituation.docx (Dec. 26, 2017).