Discrimination of Aroma Characteristics for Cubeb Berries by Sensomics Approach with Chemometrics

The dried cubeb berries are widely used as medicinal herb and spicy condiment with special flavor. However, there is a significant definition discrepancy for cubeb berries. In this study, an efficient analytical method to characterize and discriminate two popular cubeb fruits (Litsea cubeba and Piper cubeba) was established. The aroma profiles of cubeb berries were evaluated by different extraction methods including hydro-distillation, simultaneous distillation/extraction, and solid-phase micro-extraction followed by gas chromatography-mass spectrometry-olfactometry (GC-MS-O). In total, 90 volatile compounds were identified by HD, SDE, and SPME combined with GC-MS. Principal component analysis was further applied and discriminated ambiguous cubeb berries by their unique aromas: Litsea cubeba was characterized by higher level of d-limonene (“fruit, citrus”), citral (“fruit, lemon”) and dodecanoic acid; α-cubebene (“herb”) was identified as a marker compound for Piper cubeba with higher camphor (“camphoraceous”), and linalool (“flower”). Flavor fingerprint combined with PCA could be applied as a promising method for identification of cubeb fruits and quality control for food and medicinal industries.


Introduction
Litsea cubeba (Lour.) Pers. (Lauraceae) gives off an aromatic odor and smells similar to an intensely lemonlike, spicy aroma. Litsea cubeba (L. cubeba) is a promising industrial crop as its fruit is rich in valuable essential oil. Recently, many reports have demonstrated the bioactivities of essential oil in L. cubeba [1][2][3][4]. L. cubeba has been widely employed in a flavoring or herbal medicinal industries and could be used as an ingredient in ionone flavors, botanical insecticides, food spices, and personal-care products.
The dried berry of Piper cubeba (Piperaceae), known as the 'cubeb pepper' or 'tailed pepper', have been widely used as a popular spice, with beneficial properties, including anti-inflammatory, analgesic, anti-proliferative, and leishmanicidal activities [5,6], and a flavoring agent for gins and cigarettes consumed throughout Europe as well as in many other Polynesian countries [7].

Optimization of SPME
The volatile compounds in cubeb fruits were extracted using HS-SPME and the highest peak area response was selected in order to optimize the main parameters. These different desorption times, incubation times, extraction temperatures, and extraction times were optimized based on the total ion response in the GC-MS [19]. As shown in Figure 1a, study of desorption time including 1, 2, 3, and 4 min was tested. The peak areas of different volatiles were not significantly affected by the desorption time (p < 0.05). In order to clean the fiber sufficiently, 3 min was chosen as the desorption time in the present work. Figure 1b shows the effect of incubation time (10,15,20, and 25 min) on the detection of total volatiles. The total volatile amounts significantly rose with increasing incubation time. However, there was no significant difference between 15, 20, and 25 min (p < 0.05), which indicated that a 15 min incubation time would allow distributions between the fiber, the vial headspace, and the analytes to reach an equilibrium.
Study of the extraction temperature including 40, 50, 60, and 70 • C was investigated as illustrated in Figure 1c. The peak areas of volatiles were significantly affected by the extraction temperature. In order to avoid aroma changes cause by higher temperature, 60 • C was chosen as the extraction temperature.
Similarly, Figure 1d showed that there was no significant difference between 30 min and 40 min (p < 0.05). In order to avoid the fiber desorption caused by the long time exposure in the vial, 30 min was chosen as the optimum extraction time.
Therefore, the optimal extraction conditions were as follows: desorption time, 1 min; incubation time, 15 min; extraction temperature, 60 • C; and extraction time, 30 min. These conditions were applied during the headspace extraction of volatile compounds from cubeb fruits.
Molecules 2018, 23, x FOR PEER REVIEW 3 of 15 there was no significant difference between 15, 20, and 25 min (p < 0.05), which indicated that a 15 min incubation time would allow distributions between the fiber, the vial headspace, and the analytes to reach an equilibrium. Study of the extraction temperature including 40, 50, 60, and 70 °C was investigated as illustrated in Figure 1c. The peak areas of volatiles were significantly affected by the extraction temperature. In order to avoid aroma changes cause by higher temperature, 60 °C was chosen as the extraction temperature.
Similarly, Figure 1d showed that there was no significant difference between 30 min and 40 min (p < 0.05). In order to avoid the fiber desorption caused by the long time exposure in the vial, 30 min was chosen as the optimum extraction time.
Therefore, the optimal extraction conditions were as follows: desorption time, 1 min; incubation time, 15 min; extraction temperature, 60 °C; and extraction time, 30 min. These conditions were applied during the headspace extraction of volatile compounds from cubeb fruits.

Identification of Aroma Compounds
To obtain a wider volatile profile and better discriminate the two cubeb berries, three extraction methods (HD, SDE, and SPME) were used (Table 1). In total, 90 volatile compounds were identified by HD, SDE, and SPME combined with GC-MS. Seventy-three volatile compounds belonged to different chemical families: terpenes (8.82-80.65% for C. cubeba; 18.69-52.27% for P. cubeba), ketones (1.49-3.24% for C. cubeba; 17.97-20.99% for P. cubeba), alcohols (4.16-10.56 % for C. cubeba; 16.68-28.97% for P. cubeba), aldehydes (2.02-23.79% for C. cubeba; 1.71-6.43% for P. cubeba), esters (0-1.06% for C. cubeba; 0.98-2.57% for P. cubeba), and acids (2.39-48.55% for C. cubeba; 0-18.69% for P. cubeba). In order to highlight the differences between the two cubeb fruit with different extraction techniques in a simple and immediate way, Figure 2 showed the comparison of the relative percentages of the main chemical families present in cubeb berries. In order to highlight the differences between the two cubeb fruit with different extraction techniques in a simple and immediate way, Figure 2 showed the comparison of the relative percentages of the main chemical families present in cubeb berries. One of the most abundant chemical families identified in cubeb fruits was terpenes. To our knowledge, the literature dealing with the comparison of different extraction methods of volatile compounds in cubeb berries is scare. Wang et al. [20] studied the chemical composition of the essential oil obtained only by HD of different parts (root, stem, leaf, flower, and fruit) of L. cubeba and showed that citral and limonene were the main constituents. In the present work, for L. cubeba, Dlimonene was one of the main terpenes for the SDE (10.57%) and SPME extracts (38.89%), but it was not detected in the HD extract. Wang et al. [21] reported that some other terpenes, such as α-pinene, β-pinene, and β-caryophyllene, were also detected as the main volatile compounds in the bio-oils produced from L. cubeba seed by hydrothermal liquefaction. α-Pinene (5.74%), β-pinene (6.22%), Dlimonene (9.84%), and caryophyllene (5.57%) were also the main terpenes for the SPME extract of P. cubeba. γ-Terpinene (A13, 0.12-0.39%), α-copaene (A19, 0.13-1.15%), caryophyllene (A23, 5.53-9.11%), and humulene (0.77-4.33%) were the common terpenes detected in all the three different extracts of two cubeb berries. Three terpenes could only be detected in all the extracts of P. cubeb, including α-cubebene (A18, 0.4-0.52%), bicyclosesquiphellandrene (A27, 2.41-5.28%), and αfarnesene (A30, 0.13-0.16%). Cubebene was also identified in the direct analysis in real time mass spectrometry (DART-MS) fingerprint of P. cubeba studied by Kim et al. [7]. The extracts obtained by SPME were rich in terpenes (80.65% for L. cubeba and 52.27% for P. cubeba) in comparison with those from HD (8.82% for L. cubeba and 18.69% for P. cubeba) and SDE (29.9% for L. cubeba and 35.4% for P. cubeba). However, SDE with solvents tends to extract a higher amount of the volatile monoterpenes than SPME in bay leaf [22], French beans [23], and wines [24]. These differences may be due to the matrix effect in releasing volatile compounds as each spice had a characteristic plant tissue structure. Meanwhile, strong oxidation and degradation of terpenes may occur for HD and SDE extracts because of higher temperature and longer time [25]. One of the most abundant chemical families identified in cubeb fruits was terpenes. To our knowledge, the literature dealing with the comparison of different extraction methods of volatile compounds in cubeb berries is scare. Wang et al. [20] studied the chemical composition of the essential oil obtained only by HD of different parts (root, stem, leaf, flower, and fruit) of L. cubeba and showed that citral and limonene were the main constituents. In the present work, for L. cubeba, D-limonene was one of the main terpenes for the SDE (10.57%) and SPME extracts (38.89%), but it was not detected in the HD extract. Wang et al. [21] reported that some other terpenes, such as α-pinene, β-pinene, and β-caryophyllene, were also detected as the main volatile compounds in the bio-oils produced from L. cubeba seed by hydrothermal liquefaction. α-Pinene (5.74%), β-pinene (6.22%), D-limonene (9.84%), and caryophyllene (5.57%) were also the main terpenes for the SPME extract of P. cubeba. γ-Terpinene (A13, 0.12-0.39%), α-copaene (A19, 0.13-1.15%), caryophyllene (A23, 5.53-9.11%), and humulene (0.77-4.33%) were the common terpenes detected in all the three different extracts of two cubeb berries. Three terpenes could only be detected in all the extracts of P. cubeb, including α-cubebene (A18, 0.4-0.52%), bicyclosesquiphellandrene (A27, 2.41-5.28%), and α-farnesene (A30, 0.13-0.16%). Cubebene was also identified in the direct analysis in real time mass spectrometry (DART-MS) fingerprint of P. cubeba studied by Kim et al. [7]. The extracts obtained by SPME were rich in terpenes (80.65% for L. cubeba and 52.27% for P. cubeba) in comparison with those from HD (8.82% for L. cubeba and 18.69% for P. cubeba) and SDE (29.9% for L. cubeba and 35.4% for P. cubeba). However, SDE with solvents tends to extract a higher amount of the volatile monoterpenes than SPME in bay leaf [22], French beans [23], and wines [24]. These differences may be due to the matrix effect in releasing volatile compounds as each spice had a characteristic plant tissue structure. Meanwhile, strong oxidation and degradation of terpenes may occur for HD and SDE extracts because of higher temperature and longer time [25]. For ketones, 6-methyl-5-hepten-2-one (B1) and camphor (B2) were both detected in all the cubeb berries samples. As irregular terpene, 6-methyl-5-hepten-2-one is probably derivative of carotenoids produced by enzymatic action [26]. Camphor (17.59-20.6%) was detected as the major ketones in all the extracts of P. cubeba. The contents of ketones in P. cubeba (17.97-20.99%) with different extraction methods were significantly higher than in the L. cubeba (1.49-3.24%) by one-way ANOVA (Figure 2). Alcohols were also present with a high proportion in cubeb fruit, and the contents of alcohols in P. cubeba (16.68-28.97%) with different extraction methods were significantly higher than in the L. cubeba (4.16-10.56%) by a One-way ANOVA (Figure 2). Terpinen-4-ol (C7) could be detected in all the extracts of L. cubeba and P. cubeba. Linalool (C2, 14.89-21.31%) was the most abundant volatile and could be found in all the extracts of P. cubeba.
Citral was present as the most abundant aldehyde compound in L. cubeba extracts. This result is in accordance with the study by Wang et al. [20], who reported that citral was one of the main constituents in the fruit oil of L. cubeba extracted by HD. The contents of 3,7-dimethyl-2,6-octadienal in L. cubeba was higher than P. cubeba. Only 10 esters were identified in the current work, most were relatively high-boiling esters and with lower contents than other chemical families in cubeba berries.
Only two kinds of acids, decanoic acid and dodecanoic acid, were mainly extracted by HD and SDE with longer extraction time and higher temperature. The lack of acids with low volatility in SPME extracts may be caused by the low extraction temperature during the extraction process. Most acids may exist as esters form or have been changed to aldehydes, alcohols, or other secondary metabolites [13]. The contents of acids in L. cubeba (2.39-48.55%) with different extraction methods were significantly higher than in the P. cubeba (0-18.69%) by one-way ANOVA (Figure 2).

Aroma-Active Compounds by GC-MS-O
The extracts obtained by SPME were analyzed to assess the aroma-active compounds of the cubeb fruits using GC-O. Table 2 listed the identified aroma-active compounds of the L. cubeba and P. cubeba. A total of 12 compounds were tentatively found to be the aroma-active compounds at olfactometry port for odor description in GC-O analysis, including eight terpenes, one ketone, two alcohols, and one aldehyde. The odor descriptions of all the aroma-active compounds identified in the volatiles of L. cubeba and P. cubeba were basically similar to the reported of other fruits, such as blackberry [27], bayberry [17], strawberry [28], orange [29], and gooseberry [30]. According to the evaluation of the odor and the odor description of the reported, it can be concluded that the flavor of turpentine-like might be caused by α-pinene (A1), β-pinene (A4), α-phellandrene (A6), γ-terpinene (A13), and terpinen-4-ol (C7); the fruity and flower flavor might be due to the presence of D-limonene (A10), citral (D3) with higher level in L. cubeba, and linalool (C2) with higher level in P. cubeba; the herbal flavor might come from β-ocimene (A12), α-cubebene (A18) identified only in P. cubeba, and caryophyllene (A23); camphoraceous flavor might be caused by camphor (C2), which is stronger in P. cubeba with higher content than in L. cubeba. It is interesting to note that α-cubebene was in very low proportions, also had high detection frequencies in all the extracts of P. cubeba. α-cubebene had been detected in other study for cubeb fruit [7].

Principal Component Analysis (PCA)
Principal component analysis (PCA) is an unsupervised clustering method and could reduce the dimensionality of multivariate data and preserve most of the variance therein [31]. To get a clear distribution of the volatiles with the separation of the samples, PCA was applied to the data presented in Table 1, the first two principal components explained nearly 91% of the total variability of the GC-MS data set between the samples, is shown in Figure 3a. The corresponding loading weight plot, establishing the magnitude of each volatile component (variable), is illustrated in Figure 3b. Figure 3 plots the samples on the coordinate grid defined by the first two principal components and showed that PC1 and PC2 separated the L. cubeba samples from the P. cubeba samples.
Molecules 2018, 23, x FOR PEER REVIEW 9 of 13 due to the presence of D-limonene (A10), citral (D3) with higher level in L. cubeba, and linalool (C2) with higher level in P. cubeba; the herbal flavor might come from β-ocimene (A12), α-cubebene (A18) identified only in P. cubeba, and caryophyllene (A23); camphoraceous flavor might be caused by camphor (C2), which is stronger in P. cubeba with higher content than in L. cubeba. It is interesting to note that α-cubebene was in very low proportions, also had high detection frequencies in all the extracts of P. cubeba. α-cubebene had been detected in other study for cubeb fruit [7].

Principal Component Analysis (PCA)
Principal component analysis (PCA) is an unsupervised clustering method and could reduce the dimensionality of multivariate data and preserve most of the variance therein [31]. To get a clear distribution of the volatiles with the separation of the samples, PCA was applied to the data presented in Table 1, the first two principal components explained nearly 91% of the total variability of the GC-MS data set between the samples, is shown in Figure 3a. The corresponding loading weight plot, establishing the magnitude of each volatile component (variable), is illustrated in Figure 3b. Figure 3 plots the samples on the coordinate grid defined by the first two principal components and showed that PC1 and PC2 separated the L. cubeba samples from the P. cubeba samples.  Principal component 1 (PC1) and PC 2, explained 47% and 44% of the total variance among the sample batches, showed that the cubeb berries discrimination based on varietal volatile profile.

Materials and Chemicals
Litsea cubeba was collected from Guizhou province (Guiyang, China) and Piper cubeba was from Yunnan province of China (Yuxi, China). The collected cubeb samples were kept in a dry and dark place and stored at 4 • C in order to minimize any deteriorative changes to the volatile components of the cubeb berries until their processing. For the precise measurements of GC-MS-O (Agilent Technologies Inc., Santa Clara, CA, USA), cubeb fruit samples were ground to a fine powder using a grinder.

Hydro-Distillation (HD)
All the air-dried cubeb berries samples (an amount of 100 g each) were subjected to hydro-distillation using a Clevenger-type apparatus to extract essential oil using the reported methods with some modifications [20,32]. The Clevenger-type apparatus consisted of a 2000 mL glass flask, a vertical tube, a condenser, a measuring tube with stopcock, and a return tube. The return tube connected the bottom of the measuring tube to the vertical tube, which combined with the top of the condenser. The flask was filled with 1200 mL of distilled water and heated by an electric heating mantle. The extraction time was 4 h, after which no more essential oil was obtained. The vapor mixture of water-essential oil produced in the flask passed through the condenser and then the distillate was collected. The essential oil in the upper layer of the distillate was dried over anhydrous sodium sulfate (Na 2 SO 4 ) and stored at 4 • C until subsequent GC-MS analysis.

Simultaneous Distillation Extraction (SDE)
SDE was performed in a modified Lickens-Nickerson apparatus (Chrompack, Netherlands) [33]. A 25 g measure crushed air-dried cubeb berry, with 1.6g sodium chloride and 200 mL distilled water, was placed in a 500 mL flask. The sample and 40 mL of a mixture of pentane-diethyl ether (1:1 v/v) solvent placed in another flask were heated up to their boiling points and the temperature conditions were maintained for about 3 h. After cooling to ambient temperature for 10 min, the pentane-diethyl ether extract was dried over anhydrous Na 2 SO 4 . The extract was kept at 4 • C until subsequent GC-MS analysis [26].

Optimization of SPME Conditions
A SPME (Supelco, Inc., Bellefonte, PA, USA) fiber (50/30 µm divinylbenzene/carboxen/ polydimethylsiloxane; DVB/CAR/PDMS) was used for volatile extraction after the fiber had been conditioned at 270 • C for 1 h. The ground samples were passed through a 20 mesh sieve to achieve uniform particle size. A 1.5 g measure of the sieved cubeb fruits powder was placed in a 20 mL vial with a sealed cap and equilibrated in a laboratory stirrer/hot plate (model PC-420, Corning Inc. Life Science, Acton, MA, USA). Then, a stainless steel needle, housing the SPME fiber, was placed through a hole to expose the fiber in the vial [19]. Three independent extractions were done for each cubeb fruit sample.

Analysis of Volatiles by GC-MS
7890A gas chromatograph with 5975C mass spectrometer selective detector (Agilent Technologies Inc., Santa Clara, CA, USA) was used, and a DB-5 capillary column (30 m ×0 .25 mm × 0.25 µm) was applied for GC-MS. The extraction was injected into the inlet of GC-MS and desorbed at 250 • C for 3 min. The injection port was operated in splitless mode, helium (99.999%) was used as carrier gas at the flow rate of 1.2 mL/min. The initial oven temperature was 40 • C (2 min), ramped at 3 • C min −1 to 170 • C (5 min), and then ramped at 10 • C min −1 to 260 • C (5 min). Mass detector conditions were performed by EI (electronic impact) mode at 70 eV, source temperature at 230 • C, mass spectra acquisition range of 45-500 amu, scanning rate of 3.18 amu/s. The transfer line temperature was 280 • C. The volatile compounds were identified by comparing the mass spectra with mass-spectral library (NIST 2011), retention index (RI), aroma description, and matching against the published data [34]. Each extract was analyzed in triplicate. Mean data and relative standard deviation (mean ± SD) of volatiles were reported.

GC-MS-Olfactometry
GC-MS-O was performed by trained panelists on a sniffing port (Sniffer 9000, Brechbühler, Schlieren, Switzerland). The modified method described earlier by Pang et al. [18] was used in this study and the conditions of SPME and GC-MS were the same to the volatile analysis described above.
Four trained panelists take part in the detection frequency analysis (DFA) combined with GC-MS-O for identification aroma-active compounds. The panel consisted of an age from 20 to 35 years (mixed of male and female). The panelists were trained by solutions of artificial odorants and different cubeb berries samples to be familiar with the odor descriptions. In total, eight runs by GC-MS-O were conducted by four assessors (two runs for one person). The judges sniffed the effluent from the mask and recorded the time and odor characteristic of the aroma-active compounds of different cubeb berries samples. When the total detection frequencies were more than twice for the odorants perceived by two different assessors at the sniffing port, the odorants were considered potential aroma-active compounds [18,35].

Statistical Data Analysis
Significant differences for the volatile constituents among the cubeb berries were determined by one-way analysis of variance (ANOVA) using a SPSS statistics (version 20.0; SPSS, Inc., Chicago, IL, USA). The column figures in the context were plotted using Origin software (version 8.5; Northampton, MA, USA). The Unscrambler v.9.7 (CAMO AS, Trondheim, Norway) software was used for the statistical analysis (PCA) on volatiles.

Conclusions
In this study, the aroma compounds of two ambiguous cubeb berries were isolated by HD, SDE, and SPME pretreatment methods in order to fully obtain the complex aroma profiles of the cubeb berries and were analyzed by GC-MS-O combined with PCA. By GC-MS-O analysis, a total of 12 aroma-active compounds were found to play a key role in the characteristic flavor of the cubeb berries. The PCA results clearly indicated that the two ambiguous cubeb berries could be discriminated by the aroma profiles. Litsea cubeba was characterized by higher level of D-limonene, citral and dodecanoic acid; Piper cubeba was marked with α-cubebene, higher camphor, and linalool. Therefore, using the volatile profile combined with PCA is an appropriate method to discriminate the cubeb berries and assure the related product quality.