Discrimination and Identification of Aroma Profiles and Characterized Odorants in Citrus Blend Black Tea with Different Citrus Species

Citrus blend black teas are popular worldwide, due to its unique flavor and remarkable health benefits. However, the aroma characteristics, aroma profiles and key odorants of it remain to be distinguished and cognized. In this study, the aroma profiles of 12 representative samples with three different cultivars including citrus (Citrus reticulata), bergamot (Citrus bergamia), and lemon (Citrus limon) were determined by a novel approach combined head space-solid phase microextraction (HS-SPME) with comprehensive two-dimensional gas chromatography-time-of-flight mass spectrometry (GC×GC-TOFMS). A total of 348 volatile compounds, among which comprised esters (60), alkenes (55), aldehydes (45), ketones (45), alcohols (37), aromatic hydrocarbons (20), and some others were ultimately identified. The further partial least squares discrimination analysis (PLS-DA) certified obvious differences existed among the three groups with a screening result of 30 significant differential key volatile compounds. A total of 61 aroma-active compounds that mostly presented green, fresh, fruity, and sweet odors were determined in three groups with gas chromatography-olfactometry/mass spectrometry (GC-O/MS) assisted analysis. Heptanal, limonene, linalool, and trans-β-ionone were considered the fundamental odorants associated with the flavors of these teas. Comprehensive analysis showed that limonene, ethyl octanoate, copaene, ethyl butyrate (citrus), benzyl acetate, nerol (bergamot) and furfural (lemon) were determined as the characterized odorants for each type.


Introduction
Tea is now the second-most popular alcohol-free beverage worldwide and has great economic importance. Remarkably, so-called blend-tea (scented tea and flavored tea) maintains its popularity in Europe, the USA, and also nowadays, in the Asia zone for its unique flavor and remarkable health

Sensory Evaluation
Fifty samples were divided into citrus, bergamot, and lemon groups (Supporting information Table S1) by main ingredient-different Citrus species. The bergamot group (BG) comprised black tea and bergamot peel or oil, which is popular in Europe called Earl Grey. The lemon group (LG) consisted Molecules 2020, 25, 4208 3 of 15 of lemon peel with black tea, which is popular in China. The citrus group (CG) comprised the fruits of Citrus reticulata with black tea except for bergamot or lemon.
Sensory evaluation scored the representative samples as >4.0 (maximum = 8.0). Table 1 shows that 3 CG, 5 BG, and 4 LG samples had significant citrus-/lemon-like flavors. In CG, sweet, fruity and floral citrus-like were noted with a slight tea flavor, a bergamot-like and medical aroma with a slight tea flavor for BG, and a relatively heavy tea-like, clearly lemon-like and fresh flavor for LG.

Optimized Volatile Analyzing Approach Combining HS-SPME and GC×GC-TOFMS
As black tea composition is relatively large, the time of extraction is referred to as the 60 min extracted by Dr. Kang's research [14]. Then the extraction fiber, tea-water proportion, and extraction temperature methods were optimized.
Among the investigated fibers, CAR/PDMS (57335-U) had the highest component numbers and total peak area (Supporting Information Figure S1: A 1 and A 2 ). The optimal proportion of tea to water was 1:4, which resulted in significantly higher component numbers and total peak area (Supporting Information Figure S1: B 1 and B 2 ). The number of compounds did not significantly differ among extraction temperatures, but the total peak area was significantly larger at 60 • C than others (Supporting Information Figure S1: C 1 and C 2 ). Therefore, the optimized extraction conditions were CAR/PDMS fiber, tea-to-water ratio of 1:4, and extraction for 60 min at 60 • C.

Identification of Aroma Profiles in Citrus Blend Black Tea
About 800-1000 peaks were initially detected in samples with a minimum S/N ration of 50. After peak alignment, we initially identified 664 common volatile compounds by comparisons with mass spectra in the NIST 2014 library with a minimum similarity of 75%. Subsequently, the retention index (RI, the Kovats index) values of all compounds were calculated and compared with known RI to validate the accuracy of compound identification. The compounds whose RI value had a difference bigger than 20 by compared with the reported were deleted. Finally, 348 volatile compounds were ultimately identified as Reliable. These comprised 60 esters, 55 alkenes, 45 aldehydes, 45 ketones, 37 alcohols, 20 aromatic hydrocarbons, 20 oxyheterocyclic compounds, 19 nitrogen-containing compounds, 18 alkanes, 15 ethers, 7 phenols, 5 acids, 1 sulfur-compound, and 1 alkyne (Supporting Information Table S2). Moreover, some important volatile compounds were verified by comparison with standards.
The distribution of the volatiles compounds in the citrus blend black teas is shown in Figure 1. The ratio (%) of alkenes was the highest (49.77% (CG); 47.16% (BG); 41.05% (LG)). Aldehyde was the second abundant compound classification with obvious differences existed in content levels among the three groups, that the contents in LG (26.31%) was twice higher than BG (11.20%). Ester was the third larger composition in CG (17.00%) and BG (16.82%); while ester (7.19%) was near with aromatic hydrocarbon (7.33%) in LG. The ranges of alcohol and aromatic hydrocarbon were 4.85-8.86% and 5.91-10.70%, respectively. The distribution of oxyheterocyclic compounds significantly differed among the three groups, being 6.24% in LG, but low in CG (0.22%) and BG (1.02%). The ratio (%) of alkenes was the highest (49.77% (CG); 47.16% (BG); 41.05% (LG)). Aldehyde was the second abundant compound classification with obvious differences existed in content levels among the three groups, that the contents in LG (26.31%) was twice higher than BG (11.20%). Ester was the third larger composition in CG (17.00%) and BG (16.82%); while ester (7.19%) was near with aromatic hydrocarbon (7.33%) in LG. The ranges of alcohol and aromatic hydrocarbon were 4.85-8.86% and 5.91-10.70%, respectively. The distribution of oxyheterocyclic compounds significantly differed among the three groups, being 6.24% in LG, but low in CG (0.22%) and BG (1.02%).  In detail, limonene was absolutely the highest constituent (CG, 35.41%; BG, 26.10%; LG, 29.44%), followed by β-myrcene (CG, 4.75%; BG, 5.73%; LG, 4.25%). Benzaldehyde was the main aldehyde, which was reported in many famous Chinese black teas, and considered as a key aroma compound in them contributing floral odorant [17,18]. p-Cymene, a common aromatic hydrocarbon in nature, with various biological activities, was detected in a high level (CG, 2.82%; BG, 6.93%; LG, 5.80%). Linalool, which is considered the most important odorant in most black teas [7,19], was also identified in blend teas at high ratios of 1.59-4.20%.

Discrimination of Crucial Differential Volatiles in Citrus Blend Black Tea with Different Species
The aromatic profiles of three groups significantly differed. Therefore, we determined the key responsible volatiles that was considered as potentially characterized odorants of corresponding citrus blend black teas. We performed PLS-DA based on the normalized peak areas of 348 identified aroma compounds to obtain an overview of the distribution of differential volatile compounds among the three groups. The CG, BG, and LG were clearly discriminated by a PLS-DA model (R 2 Y = 0.907, Q 2 = 0.836; Figure 2A), subsequent cross-validation confirmed the reliability of the model (R 2 = 0.147, Q 2 = −0.324; Figure 2B).
We screened 30 compounds as key differential volatiles based on variable importance of projection (VIP) values in the PLS-DA model with a threshold of 1.0 and p < 0.05 (Tukey s-b(K) tests). Subsequently, specific content differences of potentially characterized odorants were elucidated using hierarchical cluster analysis (HCA). The content distribution of key volatiles could be roughly divided into three classes ( Figure 3). These classes comprised 12, 15, and 3 compounds that were obviously more abundant in CG, BG, and LG, respectively.  We screened 30 compounds as key differential volatiles based on variable importance of projection (VIP) values in the PLS-DA model with a threshold of 1.0 and p < 0.05 (Tukey s-b(K) tests). Subsequently, specific content differences of potentially characterized odorants were elucidated using hierarchical cluster analysis (HCA). The content distribution of key volatiles could be roughly divided into three classes ( Figure 3). These classes comprised 12, 15, and 3 compounds that were obviously more abundant in CG, BG, and LG, respectively.

Identification of Aroma-Active Compounds in Citrus Blend Black Teas
Active aromatic compounds were assessed in equal amounts of mixed samples from each group using GC-O/MS. Table 2 shows that at least three panelists recognized 61 active aromatic compounds (Supporting Information Tables S3-S5). By comprehensively combining the findings of panelists, we assigned 17 components to class A (fresh and green scents), 29 to class B (floral, fruity, or sweet scents), 9 to class C (herbal or woody scents), 2 to class D (bakery scents), and 4 to class E (unpleasant

Identification of Aroma-Active Compounds in Citrus Blend Black Teas
Active aromatic compounds were assessed in equal amounts of mixed samples from each group using GC-O/MS. Table 3 shows that at least three panelists recognized 61 active aromatic compounds (Supporting Information Tables S3-S5). By comprehensively combining the findings of panelists, we assigned 17 components to class A (fresh and green scents), 29 to class B (floral, fruity, or sweet scents), 9 to class C (herbal or woody scents), 2 to class D (bakery scents), and 4 to class E (unpleasant odor). The compounds in classes A and B might have directly contributed to the overall aroma quality of the citrus black teas owing to their similar scent types to sensory evaluation findings that all these samples have the obvious smell of fresh, fruity, sweet, and floral scents. Among them, heptanal (fresh, green; AI, 2.33-2.6), limonene (lemon-like, fruity, fresh; AI, 2.0-2.33), linalool (floral; AI, 2.57-3.29) and trans-β-ionone (floral; AI, 2.67-2.75) were detected in all groups.   Note: *: the compound was identified by authentic standards; [1] the classification of odor characteristics of each compounds, Class A: fresh and green scents; Class B: floral, fruity and sweet scents; Class C: herbal and wood scents; Class D: a bake scent; Class E: an unpleasant scent.
The AI was notably highest for geranyl acetate (class B; AI, 3.29), β-pinene (class A; AI, 3.17), α-pinene (class B; AI, 3.17), linalool (class B; AI, 3.14), 3-(methylthio)-nonanal (class A; AI, 3.0), most of which were in class B. In addition to the concentrated distribution of odorants in class B, more compounds in LG belonged to classes D and E, which differed with the other groups.
A radar map based on the total AI values of odorants in each class was applied to determine overall flavor profiles and differences among the three types. Class B had the most outstanding flavor attributes, although their flavor profiles and corresponding AI significantly differed among all groups (Figure 4). The ANOVA results revealed that the most discriminative attributes were in class B (CG and BG, p < 0.01; BG and LG, p < 0.05), D (CG and LG, p < 0.01; BG and LG, p < 0.01), and E (CG and BG, p < 0.05; CG and LG, p < 0.01; BG and LG, p < 0.01), whereas class A and C did not significantly differ. The CG contributed most to the class B attribute, the total intensity (41.77) of the active aromatic compounds was significantly higher than other groups (BG, 24.70; LG, 36.52), which was basically consistent with the sensory evaluation. In addition to class B, the AI were significantly higher for class D and E in LG which may be responsible for the medical odor determined in the assessment of overall flavor quality in sensory evaluation. By contrast, all attributes in BG were less pronounced than other groups, but the scores of sensory evaluations did not significantly differ, indicating lower sensitivity of the sensory evaluation.
A radar map based on the total AI values of odorants in each class was applied to determine overall flavor profiles and differences among the three types. Class B had the most outstanding flavor attributes, although their flavor profiles and corresponding AI significantly differed among all groups (Figure 4). The ANOVA results revealed that the most discriminative attributes were in class B (CG and BG, p < 0.01; BG and LG, p < 0.05), D (CG and LG, p < 0.01; BG and LG, p < 0.01), and E (CG and BG, p < 0.05; CG and LG, p < 0.01; BG and LG, p < 0.01), whereas class A and C did not significantly differ. The CG contributed most to the class B attribute, the total intensity (41.77) of the active aromatic compounds was significantly higher than other groups (BG, 24.70; LG, 36.52), which was basically consistent with the sensory evaluation. In addition to class B, the AI were significantly higher for class D and E in LG which may be responsible for the medical odor determined in the assessment of overall flavor quality in sensory evaluation. By contrast, all attributes in BG were less pronounced than other groups, but the scores of sensory evaluations did not significantly differ, indicating lower sensitivity of the sensory evaluation.

Discussion
In this study, the aroma profiles of different citrus blend black teas were investigated for the first time. Then, the aroma-active compounds were analyzed by GC-O/MS. Finally, a comprehensive conjoint analysis was made to identify the key aroma compounds in each group.

Aroma Profiles in Citrus Blend Black Tea
The ratio (%) of alkenes was the highest (49.77% (CG); 47.16% (BG); 41.05% (LG)) which agreed with previous findings of citrus [12]. Limonene and β-myrcene were the most important volatile components in orange (Citrus sinensis), lemon (Citrus limon), and mandarin (Citrus reticulata), and a very low content had also been identified in pure tea [12,13,19,20]. Compared with the previous report in lemon and tea, p-Cymene was much higher in our results [7,13,18,21]. It was supposed that p-cymene and linalool may come from both citrus and tea leaves. It might be the simple additive effects between the volatile compounds of pure tea and citrus.
The main volatile compounds identified in bergamot were also highly identified in BG [5,22]. These compounds might have been sourced from bergamot added during processing. A similar result was found in LG, in which LC and lemon had the same volatile compounds with high contents [2,13]. This might be due to the addition of citrus ingredients, resulting in a higher proportion of related citrus aroma profiles in the overall aroma composition of the blend tea. This made the citrus blend black tea contains a high content of alkenes and the resemble volatile components similar to the same cultivar citrus.
The obvious difference could be found in sensory evaluation results. The analysis of potentially characterized odorants of citrus blend black tea showed the key different volatile compounds of the three groups of blend tea samples, which might be the reason why the flavor was different among the three groups.

Aroma-Active Compounds in Citrus Blend Black Tea
In the result of GC-O/MS, the compounds belonged to class A and class B were considered as the basic odorants that contribute to the flavor of citrus blend black teas. Although the relative content of limonene was the highest, its aromatic intensity was notably moderate according to GC-O/MS, which may be due to its higher odor threshold [23]. Conversely, although the contents of heptanal (<1% ) and trans-β-ionone (<1% ) were far below than limonene, their extremely low odor thresholds of 3 µg/kg and 0.007 µg/kg, resulted in a similar AI to limonene [16]. The major source of linalool, which had low odor thresholds and high content in samples, was difficult to determine because it is a key volatile compound in both black tea and citrus [14,24].
Terpinen-4-ol, fresh aroma, has been detected both in bergamot and tea, which was consistent with our results [5,16]. Differently, linalool, which was reported as a key volatile in Earl grey black tea (bergamot black tea) [25], showed unobtrusive AI value (2.57) comparing with those in other groups, the difference of extraction methods might have caused this.

Comprehensive Understanding of Characterized Odorants in Citrus Blend Black Tea
The characterized odorants contributing to the aroma characteristics of citrus blend black teas were less rigorous when determined by simple quantitative or olfactory analyses due to the odor characteristics of different volatiles and subjective factors associated with panelists. Moreover, some of the GC×GC-TOFMS and GC-O/MS results did not always correspond, which was probably due to slight differences in the experimental conditions and error factors. Therefore, the combined results provided a more objective and precise identification of the characterized odorants in citrus blend black teas.
The distribution trends of the contents and AI values of seven compounds (CG (4), BG (2), LG (1)) were similar among the three groups, indicating their importance to the corresponding overall aroma quality ( Figure 5). We considered that limonene (lemon-like, fruity, fresh), octanoic acid ethyl ester (green, waxy), copaene (sweet, floral), and ethyl butyrate (fruity) were the characterized odorants in CG. In fact, limonene was the most abundant and moderately intense in all samples, but the corresponding values in CG were significantly higher, indicating the superior distribution of the compounds in some citrus varieties. The other three odorants were essentially undetectable (0.39-3.64% ) in BG and LG. Their remarkably high distribution and scents may be responsible for the more intense floral and fruity scents in CG. Benzyl acetate (herbal, sweet) and nerol (floral, sweet) were determined as the characterized odorants in BG, which were not only detected in GC-O/MS but also showed more abundant. Nerol, which may be a unique volatile in BG, has previously been detected in Earl Grey tea, but its role has not been discussed in-depth, which may be due to differences in experimental design and detection methods [11,25]. Benzyl acetate might contribute to the unique herbal, medical, and sweet scents in BG. Similarly, furfural with a roasted odor was considered as one of the characterized odorants of LG.
intense floral and fruity scents in CG. Benzyl acetate (herbal, sweet) and nerol (floral, sweet) were determined as the characterized odorants in BG, which were not only detected in GC-O/MS but also showed more abundant. Nerol, which may be a unique volatile in BG, has previously been detected in Earl Grey tea, but its role has not been discussed in-depth, which may be due to differences in experimental design and detection methods [11,25]. Benzyl acetate might contribute to the unique herbal, medical, and sweet scents in BG. Similarly, furfural with a roasted odor was considered as one of the characterized odorants of LG. LG: Lemon Group.

Citrus Blend Black Tea Samples
A total of 12 representative citrus blend black tea samples including 3 CG samples, 5 BG samples and 4 LG samples were selected from 50 commercially available samples purchased at origin countries' markets. Table 1 shows the brand names and blended ingredients. All samples were fully powdered (~200 mesh using a Tube Mill 100 control grinder (IKA Werke GmbH & Co. KG, Staufen, Germany) at 5000 rpm for 20 s.

Citrus Blend Black Tea Samples
A total of 12 representative citrus blend black tea samples including 3 CG samples, 5 BG samples and 4 LG samples were selected from 50 commercially available samples purchased at origin countries' markets. Table 1 shows the brand names and blended ingredients. All samples were fully powdered (~200 mesh using a Tube Mill 100 control grinder (IKA Werke GmbH & Co. KG, Staufen, Germany) at 5000 rpm for 20 s.

Optimization of Volatile Extraction from Citrus Blend Black Teas Using HS-SPME (for GC×GC-TOFMS and GC-O/MS Analyses)
Using a multiple-factor orthogonal experiment to determine the appropriate extraction fibers among CAR/PDMS (57335-U, 85 µm), PDMS/DVB (57327-U, 65 µm), DVB/CAR/PDMS (57329-U, 50/30 µm) and PDMS (57301, 100 µm), the experiment with water-sample ratio and temperature proceeded as follows. The fibers were conditioned at high temperature (300 • C for CAR/PDMS, 250 • C for PDMS/DVB and PDMS, 270 • C for DVB/CAR/PDMS) for 0.5 h before their first use and then screened under the same HS-SPME conditions. Triplicate powdered samples (1.0 g) were each placed in 20-mL glass vials, then boiling water (2, 3, 4, 6, or 8 mL) was added. The vials were immediately placed in a heating oscillator to equilibrate for 3.0 min at 30 • C, 50 • C, 60 • C, 70 • C, 90 • C, respectively, then the solid-phase microextracted fibers (SPME) were exposed to the vial headspace, and stirred at a constant speed and temperature for 60 min. Finally, the SPME fibers were loaded into the GC×GC injector and left for 5.0 min to permit thermal desorption of the aroma extract.

GC×GC-TOFMS Analysis
GC×GC conditions: The temperature of the GC injector and the transfer line was set to 250 • C. Helium (99.999%) was the carrier gas at a constant flow of 1.0 mL/min. A split injection was applied at a split ratio of 20:1. Standards in ethanol were injected using an MPS-2 multi-purpose sampler with an injection volume of 1.0 µL, and the aroma extracts were injected using the HS-SPME auto sampling system. The temperature programs were as follows: hold at 50 • C for 2 min, increase in 8 • C/min increments to 265 • C, then hold for 5.0 min for the 1-d column, and hold at 55 • C for 2 min, increase in 8 • C/min increments to 270 • C, the hold for 5 min for the 2-d column. The modulation period was set at 5 s. TOFMS conditions: The TOFMS parameters were electron ionization at −70 eV, an ion source temperature of 220 • C, an electron multiplier at 1400 V and a mass range of 33-600 u.

GC-O/MS Analysis
GC-MS conditions: HP-5MS column, 30 m × 250 µm × 0.25 µm; GC injector temperature, 250 • C; helium (99.999%) flow, 1.6 mL/min; splitless injection and 0.0 s of modulation. The temperature program for the GC column proceeded as follows: hold at 50 • C for 3 min, increase in 4 • C/min increments to 265 • C, and then hold for 5 min. The temperature of the transfer line was 270 • C for the entire 60.75-min duration of the analysis. Mass spectrometry proceeded under an ion source temperature of 220 • C with a mass range of 33-600 u, detector voltage, 1300 V and electron ionization −70 eV.
Three male and four female panelists who were selected and trained as we have previously described, tested the aroma-active compounds in each sample three times using GC-O [26]. The intensity of each aroma was defined on a scale of 1 to 4 as weak (1), moderate (2), strong (3), and extremely strong (4) [16,27]. A general description of odorants with the same retention time by at least three panelists was selected and further determined by GC-MS and standards. The panelist scores were averaged to define the corresponding aroma intensity.

Sensory Evaluation
Two male and three female healthy tea-tasters conducted the sensory evaluation. All were selected and trained as we have previously described [14]. They had been certified by the China Tea Science Society after passing theory and practical examinations and had at least two years of experience in tea sensory evaluation.
The aroma characteristics of the samples were described according to the national standards in "Black tea" (GB/T 13738-2017) and "Teabag" (GB/T 24690-2018). The aroma qualities of samples were separated into two parts. The tea-tasters should score the intensity of the tea and citrus aromas that were smelled in each sample, respectively. Scores were based on the aroma intensity (AI) method in GC-O with the modification of adding a score of 1 to 4 indicate the absence of a tea or citrus fragrance, and a score of 0 meant that there was no such scent. The final score was the sum of the averaged dimensional intensities determined by the tasters.

Data Processing
The GC×GC-TOFMS data preprocessing method in the LECO Chroma TOF software was used. Partial least squares discriminant analysis (PLS-DA) using the SIMCA-P 12.0 software (Umetrics Corporation, Umeå, Sweden) and hierarchical clustering analysis (HCA) using the MultiExperiment Viewer 4.8.1 (Oracle Corporation, Redwood Shores, California CA, USA) were performed to the statistical multivariate analyses. ANOVA analysis was performed using SPSS Statistics 20.0 (IBM Corp., Armonk, NY, USA).

Conclusions
In conclusion, the aroma profile and characteristic odorants in the main current commercial citrus blend black teas were thoroughly investigated utilizing HS-SPME-GC×GC-TOFMS combined with GC-O/MS techniques for the first time. A total of 348 volatile compounds were ultimately identified that consisted of >50% alkenes and aldehydes. The PLS-DA resulted in 30 significant differential volatile compounds among the three types. Moreover, GC-O/MS analysis revealed 61 aroma-active compounds. Most of these compounds presented green, fresh, floral, lemon-like fruity, and sweet scents. Especially heptanal, limonene, linalool, and trans-β-ionone were considered as the basic odorants for citrus blend black tea flavors. The combination of GC×GC-TOFMS and GC-O/MS indicated that each group had a unique composition of volatile compounds, namely, limonene, octanoic acid ethyl ester, copaene, ethyl butyrate in CG, benzyl acetate, and nerol in BG, and furfural in LG. The combination of results reliably identified blended tea aromas and flavors and led to a comprehensive understanding of the flavor sources in citrus blend black tea. Meaningful, the result also lays the foundation for the cultivar discrimination and aroma quality control of the popular blended teas. Our subsequent studies will focus on identifying unknown odorants, improving analytical approaches, and uncovering synergistic and inhibitory effects among odorants.