Effect of Drying Methods on Volatile Compounds of Burdock (Arctium lappa L.) Root Tea as Revealed by Gas Chromatography Mass Spectrometry-Based Metabolomics

Burdock (Arctium lappa L.) is one of the nutritional foods widely planted in many countries. Dried burdock root (BR) is available as a herbal tincture and tea in many Asian countries with good flavor and taste. In this study, the volatile components in dried BR were identified and the effects of different drying methods on the volatile components were investigated by HS-GC-MS method. A total of 49 compounds were identified. Different drying methods including hot-air drying (HD, at 50, 60, 70, and 80 °C), vacuum drying (VD, at 50, 60, 70, and 80 °C), sunlight drying (SD), natural drying (ND), and vacuum freeze drying (VFD) were evaluated by HS-GC-MS-based metabolomics method. Results showed that different drying methods produced different effects on the volatile compounds. It was observed that 2,3-pentanedione, 1-(1H-pyrrol-2-yl)-ethanone, furfural, and heptanal were detected at higher concentrations in HD 80 and VD 70. The traditional HD and SD methods produced more flavor substances than VFD. The BR treated by the VFD method could maintain the shape of the fresh BR pieces while HD50 and VD80 methods could maintain the color of fresh BR pieces. These findings could help better understand the flavor of the corresponding processed BR and provide a guide for the drying and processing of BR tea.


Introduction
Arctium lappa L. (burdock), which is also called "Niubang" in China, is a biennial herb of the compositae family. It is not only a nutritious vegetable but also an important traditional Chinese medicine (TCM) that possesses antioxidant [1][2][3], antihyperglycemic [4], high fat diet-induced quail atherosclerosis protective effects [5], and other pharmacological effects [6]. It was reported that phenolic and polysaccharide were the main compounds [7,8]. In addition, many of the biological properties can be attributed to its phenolic compounds such as caffeoylquinic acids, chlorogenic acids, and their derivatives [9][10][11]. In addition to being consumed fresh or as a TCM ingredient, burdock root (BR) is primarily dried into tea in China and Southeast Asia. The aroma that is determined by a group of volatile organic compounds (VOCs) is an important factor of the quality of dried tea [12]. Drying is an essential step in the processing of dried burdock tea and the drying methods will have a great impact on the VOCs and flavor [13,14]. For example, high temperature drying and lyophilization caused the highest essential oil loss of Thymus vulgaris as analyzed by GC-MS [15]. Ye et al. [14] studied the effects of different drying methods on the aroma compounds of Anoectochilus roxburghii (Wall.) Lindl. and found that vacuum drying gave best results. Although studies about the separation and identification of nonvolatile compounds from BR have been performed, little is known about the VOCs in dried BR tea and the effects of different drying methods on it.
Many methods have been developed and applied for the analysis of chemical components from burdock [11,16,17]. The composition of volatile components from burdock leaves has been determined by gas chromatography-mass spectrometry (GC-MS) [17]. In recent years, the GC-MS-based metabolomics method has been developed for the study of the aroma in food products and processing [18,19]. The GC-MS-based metabolomics method is very useful for comprehensive and systematic analysis of VOCs in biological samples, including alcohols, hydroxy acids, amino acids, fatty acids, sterols, etc. [20]. Xu et al. [21] applied an untargeted GC-MS-based plant metabolomics method for comparing volatile oil profiles of Mahuang and Mahuanggen. To the best of our knowledge, the GC-MS-based metabolomics approach has not yet been applied to analyze the effects of different drying methods of the VOCs in dried BR tea.
In this work, the volatile compounds from the dried BR were investigated by headspace (HS)-GC-MS. Different drying treatments, including hot-air drying (HD, at 50, 60, 70, and 80 • C), sun drying (SD), vacuum freeze drying (VFD), vacuum drying (VD, at 50, 60, 70, and 80 • C), and natural drying (ND) were prepared and analyzed by the HS-GC-MS method. The multivariate statistical analysis including the orthogonal partial least squares discriminant analysis (OPLS-DA) and hierarchical cluster analysis (HCA) were applied to investigate the potential key discriminate compounds between ND and other drying methods.

Materials
The fresh BR was collected from Linyi City, Shandong Province (35 • 05 N, 118 • 24 E) in September 2019, and these samples were identified based on morphological features by the lecturer Ai-cun Zhou from Zhejiang A & F University. The BR samples were stored in a refrigerator at 4 • C until use. These samples were rinsed and cut into small pieces (about 3 mm thick) before drying. The reference standards C 7 -C 40 saturated n-alkane mixture (1000 mg/L) was purchased from Sigma-Aldrich Co., Ltd. (Shanghai, China).

Sample Preparation 2.2.1. Hot-Air Drying (HD) Procedure
About 60 g of freshly cut BR pieces were spread on the watch glasses, which were then placed in a drying oven (Shanghai Jinghong Experimental Equipment Co., Ltd., Shanghai, China). The samples were then dried to constant weight at 50, 60, 70, and 80 • C, respectively.

Sun Drying (SD) Procedure
About 60 g of freshly cut BR pieces were spread on the watch glasses, which were put into a drug illuminate chamber (Shanghai Jianheng Instrument Co., Ltd., Shanghai, China). The samples were then dried to constant weight at 5900~6650 Lux.

Vacuum Freeze-Drying (VFD) Procedure
About 60 g of freshly cut BR pieces were firstly frozen in an ultra-low temperature refrigerator (Thermo Scientific Forma 900 Series, American) at −80 • C for 4 h, and then were quickly placed into a freeze dryer with a cold-trap temperature of −80 • C and pressure of 3.4 Pa (FD-80, Beijing Boyikang Lab Instrument Co., Ltd., Beijing, China) for 48 h to achieve constant weight.

Vacuum Drying (VD) Procedure
About 60 g of freshly cut BR pieces were spread on the watch glasses, which was put into a vacuum drying oven (DZF-6050, Shanghai Jinghong Experimental Equipment Co., Ltd., Shanghai, China). The samples were dried to constant weight at −0.01 MPa with temperature set at 50, 60, 70, and 80 • C, respectively.

Natural Drying (ND) Procedure
About 60 g of freshly cut BR pieces were spread on the watch glasses and then were dried indoors to allow the moisture to evaporate naturally. The average room temperature and relative humidity were 20 ± 2 • C and 40%, respectively.
All the above samples for each drying method were prepared in triplicate. The biomass density of the BR pieces was around 0.36 g/cm 2 .

Color Measurement
CIELAB color space was used for the color characterization. The parameters L* (lightness), a* (balance between green and red), and b* (balance between yellow and blue) were measured by a Chroma Meter CR-400 colorimeter (Hangzhou Kesheng Instrument Co., Ltd., Hangzhou, China). Measurement was performed on random surface of the BR samples processed by different drying methods and the fresh one. Each group of samples was determined in triplicate and the values were expressed as mean ± SD. The total color differences (∆E) were calculated by the following formula [22]: where the L 0 *, a 0 *, and b 0 * referred to the parameters of fresh BR.

Water Activity (Aw) Determination
A dew point water activity meter analyzer (Aqualab 4TEV) was used to determine the water activity of all the dried BR samples processed by different drying methods. A random piece of dried BR sample was put into the water activity analyzer, and 20 min later, the readings were recorded.

Preparation of BR Powder
The dried BR pieces were ground into powder on a Retsch PM200 planetary ball mill (Retsch Inc., Haan, Germany). The ball milling parameters were: the rotating speed was 300 rpm and the rotating time was 15 min. The filling degree was 10.47%. The powder samples were kept in a refrigerator at 4 • C before analysis.

Preparation of QC Samples
The QC (quality control) samples were prepared by mixing equal amounts (1.15 g) of the BR powder obtained from different drying methods, and then analyzed in the same way as analytic samples [23,24].

HS-GC-MS Conditions
The HS-GC-MS analysis was carried out in an Agilent 7890B-5977B gas chromatographmass spectrometer equipped with an Agilent 7697A Headspace auto-sampler (Santa Clara, CA, USA). A J&W capillary column DB-624 UI (6% Cyanopropylphenyl, 94% dimethylsiloxane) of 30 m × 0.250 mm with 1.4 µm film (USA) was used for the separation. About 1.0 g of sample powder from each drying condition and QC samples were accurately weighed into a 20 mL sealed headspace vial and analyzed. Samples of each drying method were prepared in sextuplicate.
HS parameters were as follows: The BR samples in 20 mL headspace vial were heated at 120 • C with an equilibration time of 30 min. The temperatures of the headspace loop and the transfer line were maintained at 130 and 140 • C, respectively. The gas phase was injected into the GC-MS for analysis with the press equilibration time of 0.1 min and the injection time 1.0 min. A low shaker mode of the headspace vial was set up. GC parameters were as follows: The carrier gas and make-up gas were high purity helium (>99.9%) and nitrogen (>99.9%), respectively. The carrier gas helium was used at a constant flow of 1.0 mL/min. The injector port was kept at 200 • C in split injection mode with split ratio 10:1. The oven temperature was ramped from 40 • C (held 5 min) to 170 • C at 8 • C/min, then to 230 • C (held 8 min) at 20 • C/min. The total run time was 32 min. All the samples along with QCs were injected into the GC-MS in a randomized order (QCs at regular intervals) to ensure the reliability of data. Additionally, the QC samples were injected at the beginning and at the end of the whole analytical run [21].
MS parameters were as follows: Data were acquired in electron impact (EI) mode at 70 eV, using the full scan mode over the mass range m/z 45 to 600 at 2.6 spectra/s. Temperatures of the ion source, the quadrupole, and the interface were set at 230, 150, and 250 • C, respectively. Data were analyzed by using Agilent MassHunter Qualitative Analysis (B.07.00). The identification of the VOCs was achieved by searching in the NIST14 spectral library by means of Automated Mass Spectral Deconvolution and Identification System (AMDIS) v2.72 as well as comparison of their GC retention time (RT), retention index (RI), and mass spectra with that of the saturated n-alkanes. The RI of the compound was calculated using C 7 -C 40 n-alkanes, and the RT of the n-alkane was obtained with using the same HS-GC-MS conditions as BR samples.

Data Processing
The GC-MS raw data files were firstly converted into the Analysis Base File (ABF) format by AbfConverter (version 4.0.0) and then further processed by MS-DIAL (version 4.16) [25]. Data processing included data collection (mass scan range at 45-600 Da), peak detection, deconvolution, filtering (peak count filter was set as 8.1%, as there were 6 biological replicates and the total number of data was 74), peak alignment, and integration. An integrated data matrix composed of average RT, average RI, EI spectrum, sample name, and corresponding peak area was generated. Then, the peak area from the data matrix was normalized using the locally weighted scatterplot smoothing (LOWESS) method prior to further statistical analysis.

Statistical Analysis
The processed data was introduced into SIMCA 14.1 (Umetrics, Sweden) for multivariate statistical analyses, respectively. All variables were pareto-scaled before chemometric analysis [21]. The orthogonal partial least squares discriminant analysis (OPLS-DA) was utilized to study the differences between the SD, VFD, HD, VD, and ND samples. The differential volatile components were selected according to the variable importance in projection (VIP > 1.0) obtained from the OPLS-DA model and p-values (p < 0.05) calculated by Mann-Whitney U test. Meanwhile, fold changes (FC), calculated by the mean content of each compound, were applied to express the relative content changes of different treatments. The processed data were subject to log2 transformation. The hierarchical cluster analysis (HCA) with unsupervised method by R was used to reveal the aroma compound profiling changes among all the 11 samples of different drying methods. The heat map was obtained by using the R-4.0 language p heatmap package.

Appearance of Dried Burdock
The changes of dry weight for BR under different drying methods are shown in (Figure 1). The dry weight of burdock slices obtained under the 11 different drying methods ranged from 11.1 to 12.8 g (from 60 g fresh burdock) with little differences. So, the unbound water content in fresh BR that can be removed by drying is around 80.1 ± 0.4%. The result is in accordance with the moisture content of BR reported from the literature [26]. The water activity (Aw) of fresh BR is about 0.99, which is easy to spoil during storage. The Aw of BR under hot air drying, vacuum drying, vacuum freeze drying, natrual drying, and  [27]. Among them, the vacuum freeze-dried BR sample has the lowest water activity. ranged from 11.1 to 12.8 g (from 60 g fresh burdock) with little differences. So, the unbound water content in fresh BR that can be removed by drying is around 80.1 ± 0.4%. The result is in accordance with the moisture content of BR reported from the literature [26]. The water activity (Aw) of fresh BR is about 0.99, which is easy to spoil during storage. The Aw of BR under hot air drying, vacuum drying, vacuum freeze drying, natrual drying, and sun drying are 0.3782-0.4442, 0.4027-0.5093, 0.2239, 0.3903, and 0.4498, which are all below 0.6 [27]. Among them, the vacuum freeze-dried BR sample has the lowest water activity.

Figure 1.
Appearance and dry weight of dried burdock root treated by different drying methods. (n = 3; the fresh weight for burdock is 60 g, and the dry weight are expressed as Mean ± SD; HD50-80 means hot air drying at 50-80 °C, VD50-80 means vacuum drying at 50-80 °C, VFD means vacuum freeze drying, ND means natural drying, SD means sun light drying, FP means fresh pieces and Aw means activity water).
The appearance of the dried burdock tea varied depending on the drying method used as shown in (Figure 1). The chroma data including the a*, b*, L*, and ΔE values were listed in (Table 1). The BR samples that underwent most drying methods were shrunken. However, the BR treated by the VFD method was white and with almost no shrinkage. VFD method could maintain the shape of fresh BR pieces. Compared with the fresh BR samples, no obvious decrease of L* values was observed except VD60. Besides, SD gave the highest L* value. The highest ΔE values were observed in the VD60 treated samples while the color of HD50 and VD80 methods were closest to that of fresh BR with the ΔE Figure 1. Appearance and dry weight of dried burdock root treated by different drying methods. (n = 3; the fresh weight for burdock is 60 g, and the dry weight are expressed as Mean ± SD; HD50-80 means hot air drying at 50-80 • C, VD50-80 means vacuum drying at 50-80 • C, VFD means vacuum freeze drying, ND means natural drying, SD means sun light drying, FP means fresh pieces and Aw means activity water).
The appearance of the dried burdock tea varied depending on the drying method used as shown in (Figure 1). The chroma data including the a*, b*, L*, and ∆E values were listed in (Table 1). The BR samples that underwent most drying methods were shrunken. However, the BR treated by the VFD method was white and with almost no shrinkage. VFD method could maintain the shape of fresh BR pieces. Compared with the fresh BR samples, no obvious decrease of L* values was observed except VD60. Besides, SD gave the highest L* value. The highest ∆E values were observed in the VD60 treated samples while the color of HD50 and VD80 methods were closest to that of fresh BR with the ∆E values at 4.68 and 4.82, respectively. Thus, these two drying methods were most able to maintain the color of fresh plants.

Optimization of HS-GC-MS Parameters
The equilibrium temperature was the key factor affecting the response of volatile compounds in HS [28]. To obtain the optimal equilibrium temperature, different temperatures of 100, 110, 120, 130, and 140 • C were performed in HS analysis. With the increase of equilibrium temperature, the peak response will increase. However, the extreme high temperatures will lead to the degradation and transformation of thermally unstable compounds. Therefore, 120 • C was chosen as the preferred equilibrium temperature in this work.
Additionally, the column was also one of the vital conditions for chromatographic separation. Better chromatographic separation was a prerequisite for the analysis of compounds. In this study, two capillary columns including the HP-5 MS UI (5%-phenyl)-methyl polysiloxane of 30 m × 0.250 mm with 0.25 µm film (USA) and DB-624 UI (6% Cyanopropylphenyl, 94% dimethylsiloxane) of 30 m × 0.250 mm with 1.4 µm film (USA) were evaluated for their separating ability. The preferred resolution and better peak response were observed by using the DB-624 UI capillary column.
Other HS-GC-MS conditions including the equilibrium time of HS, the injector port temperature and the oven temperature of GC, the ion source temperature and quadrupole temperature of MS, and so on, were also optimized based on the experimental data. The representative GC-MS total ion chromatogram (TIC) of BR under the optimized conditions is shown in (Figure 2).

Volatile Compounds Characterization by HS-GC-MS.
Volatile compounds were identified from the QC sample based on the average RT, calculated RI value, and mass spectrum obtained after data processing. To calculate the

Volatile Compounds Characterization by HS-GC-MS
Volatile compounds were identified from the QC sample based on the average RT, calculated RI value, and mass spectrum obtained after data processing. To calculate the RI of the matched compounds, an n-alkane mixture was run after the sample sequence. The identification criteria of a compound was the match score in the NIST greater than 750, or the difference between the calculated RI and that given from the NIST14 library was less than 50. RI was widely applied to identify compounds. Some structurally similar compounds with similar mass spectra can be distinguished due to their different RI [29]. The formulae for calculating RI value was as follows: where TR(x), TR(z), and TR(z+1) represent the retention time of the component, the nalkane with carbon number at z and z + 1, respectively, and TR(z) < TR(x) < TR(z + 1).
In the study, a total of 49 volatile compounds were identified from BR ( Table 2). These compounds mainly comprised of 11 aldehydes, 7 ketones, 9 heterocycles, 6 terpenes, 4 aromatics, 3 halogen-substituted hydrocarbons, 2 alcohols, 2 acids, 2 aliphatic hydrocarbons, and 3 other compounds. Aldehydes were one of the main volatile components in the dried BR. Compounds 4, 10, 12, 13, 17, 21, 23, 26, 31, 36, and 39 were identified as aldehydes. It is known that aldehydes are the main components of essential oils from many plants which have high antimicrobial activities and valuable applications [30][31][32]. Among them, 3-methylbutanal (compound 12), 2-methylbutanal (compound 13), hexanal (compound 21), and furfural (compound 23) were found to be the main components. Take compound 21 as an example to explain the identification work of these aldehydes. The mass spectrum matching result showed that hexanal had a high matching score of 954. The calculated RI of compound 21 was 830, which was similar with the database retention index of 800. As a result, compound 21 was identified as hexanal derived from oxidative decomposition of fatty acids. Thermal oxidation of cholesterol and Maillard reaction can lead to development of 2-methylbutanal and 3-methylbutanal. Furfural also has a role as a Maillard reaction product and a metabolite. It is a member of furans and an aldehyde which can derive from furan. Compounds 9,15,22,24,27,29,33,38, and 40 were identified as heterocycles. Compound 29 with relatively high content in the dried BR was used as an example to elucidate the identification work of these heterocycles. The mass spectrum matching result showed that 2-pentylfuran had a high matching score of 951. The calculated RI of compound 29 was 1007, which was similar with the library RI of 993 and thus the compound 29 was identified as 2-pentylfuran.

Comparisons of SD, VFD, HD, VD, and ND on the Volatile Compounds of BR by OPLS-DA
To further explore the differential compounds among the different drying methods, the OPLS-DA for pairwise comparison was carried out. Table 3 listed the overall view of FC values between different drying methods and ND. More volatile substances were observed in HD drying methods than in VD due to the oxidization or degradation of components at high temperatures. For HD80, 2,3-Pentanedione was found to be in high content. Formic acid and 2-ethylfuran were observed in all the groups. It can be seen that compared with ND, the proportion of formic acid in the other drying methods has all increased. However, the 2-ethylfuran decreased as the FC values were below 1. Besides, the hexanal, 2-heptanone, and 2-pentylfuran also decreased in most of the drying methods. The smallest FC values were observed in VFD. This result was in accordance with the findings for volatiles after VFD in coffee beans [33] and Chinese ginger [34] that VFD could not produce volatile flavor substances from samples. The volatile compounds that did not meet the identification criterion were noted as unknown.

Comparison between SD and ND
The different effects on VOCs of BR between SD and ND were displayed in ( Figure 3A) and the corresponding loading plots are shown in ( Figure S1A). Table S1 lists the main 19 volatile components identified based on VIP > 1 and p-value < 0.05. Among them, dimethyl sulfide had the highest VIP value, followed by 2-methylbutanal and 3-methylbutanal. The relative contents of aldehydes (3-methylbutanal, 2-methylbutanal, pentanal, propanal, and furfural) and terpenes (β-elemene, β-Selinene, and α-Selinene) were observed more in SD than ND. Terpenes are a series of bioactive compounds which have anticancer [35,36] and antifungal [37,38] activities. The terpenes identified in this work were all sesquiterpenes (C15) that could be synthesized through the MVA pathway [39]. Another six differential heterocyclic compounds were found to be the important components of plant flavor and the main components of heated-treated food, especially pyrazines. Besides, formic acid, chloromethane, and dimethyl sulfide were also the potential chemical markers for differentiating SD from ND.

Comparison between VFD and ND
VFD was considered as the forefront of drying research as it could maintain the shape of the final products [40] as well as good quality and nutrition of food [41], however, it often needs high energy consumption and cannot preserve the aroma substances.
Separation between VFD and ND was shown in ( Figure 3B) with high value of R2Y and Q2. As shown in (Table S2), the relative percentage of aldehydes exhibited more in VFD than in ND. Besides, the differential aliphatic hydrocarbons and two ketones (2-butanone and 2-heptanone) between VFD and ND were also observed. Aliphatic hydrocarbon compounds are derived from oxidative decomposition of lipids [42]. In this research, three differential aliphatic hydrocarbons were identified, 1-pentadecene, heptadeca-1,8,11-triene, and heptadeca-1,8,11,14-tetraene. They have relatively high odor threshold values and usually lack substantial impact on flavor [43]. Besides, another differential compound with VIP 1.59 was identified as the terpene1-methyl-4-(6-methylhept-5-en-2yl) benzene, which has the antimicrobial and antioxidant activities [44]. The loading plots between VFD and ND are shown in Figure S1B.

Comparison between HD and ND
HD is one of the most conventional drying processes widely used in the food industry [45] with the advantages of easy-operation and low cost [46]. However, it will also take away part of the volatile substances while performing the drying process. The effect of temperature on food is ultimately determined by its effect on chemical components as high temperature usually results in the degradation and loss of important flavor substances.
In this work, comparisons of HD (named HD50, HD60, HD70, and HD80 at 50, 60, 70, and 80 °C , respectively) and ND were performed. Four OPLS-DA models with distinct separations are shown in Figure 4. All the models showed good fitness and strong predictability with high value of R2Y and Q2. Afterwards, the key differential compounds of four sets of comparison selected according to VIP > 1 and p < 0.05 were displayed in (Tables S3-S6), respectively. The labels of ID were exported after peak alignment by MS-

Comparison between VFD and ND
VFD was considered as the forefront of drying research as it could maintain the shape of the final products [40] as well as good quality and nutrition of food [41], however, it often needs high energy consumption and cannot preserve the aroma substances.
Separation between VFD and ND was shown in ( Figure 3B) with high value of R2Y and Q2. As shown in (Table S2), the relative percentage of aldehydes exhibited more in VFD than in ND. Besides, the differential aliphatic hydrocarbons and two ketones (2-butanone and 2-heptanone) between VFD and ND were also observed. Aliphatic hydrocarbon compounds are derived from oxidative decomposition of lipids [42]. In this research, three differential aliphatic hydrocarbons were identified, 1-pentadecene, heptadeca-1,8,11-triene, and heptadeca-1,8,11,14-tetraene. They have relatively high odor threshold values and usually lack substantial impact on flavor [43]. Besides, another differential compound with VIP 1.59 was identified as the terpene1-methyl-4-(6-methylhept-5-en-2-yl) benzene, which has the antimicrobial and antioxidant activities [44]. The loading plots between VFD and ND are shown in Figure S1B.

Comparison between HD and ND
HD is one of the most conventional drying processes widely used in the food industry [45] with the advantages of easy-operation and low cost [46]. However, it will also take away part of the volatile substances while performing the drying process. The effect of temperature on food is ultimately determined by its effect on chemical components as high temperature usually results in the degradation and loss of important flavor substances.
In this work, comparisons of HD (named HD50, HD60, HD70, and HD80 at 50, 60, 70, and 80 • C, respectively) and ND were performed. Four OPLS-DA models with distinct separations are shown in Figure 4. All the models showed good fitness and strong predictability with high value of R2Y and Q2. Afterwards, the key differential compounds of four sets of comparison selected according to VIP > 1 and p < 0.05 were displayed in (Tables S3-S6), respectively. The labels of ID were exported after peak alignment by MS-DIAL. Compounds that did not meet the identification criteria were recorded as unknown. A total of 27, 21, 25, and 30 compounds were screened from the four groups HD50-ND, HD60-ND, HD70-ND, and HD80-ND, respectively. The corresponding loading plots are shown in (Figure S2).

Comparison between VD and ND
VD samples dried under negative pressure, reducing the processing time and allowing less oxygen exposure time [47]. Different temperatures of VD could have different effects. High temperature may lead to the degradation of biologically active components in the plant.
The supervised OPLS-DA was performed to understand the volatile compounds contributing to differentiation. In the OPLS-DA score plots ( Figure 5), a group of clear discrimination was observed between different temperatures of VD (named VD50, VD60, VD70, and VD80 at 50, 60, 70, and 80 °C , respectively). The models were stable and reproducible. A total of 21, 12, 9, and 14 compounds in VD50, VD60, VD70, and VD80 groups were selected and identified based on the criteria of VIP > 1 and p < 0.05 (Tables S7-S10), respectively. The corresponding loading plots are shown in ( Figure S3).

Comparison between VD and ND
VD samples dried under negative pressure, reducing the processing time and allowing less oxygen exposure time [47]. Different temperatures of VD could have different effects. High temperature may lead to the degradation of biologically active components in the plant.
The supervised OPLS-DA was performed to understand the volatile compounds contributing to differentiation. In the OPLS-DA score plots ( Figure 5), a group of clear discrimination was observed between different temperatures of VD (named VD50, VD60, VD70, and VD80 at 50, 60, 70, and 80 • C, respectively). The models were stable and reproducible. A total of 21, 12, 9, and 14 compounds in VD50, VD60, VD70, and VD80 groups were selected and identified based on the criteria of VIP > 1 and p < 0.05 (Tables S7-S10), respectively. The corresponding loading plots are shown in ( Figure S3).

HCA of Volatile Compounds in BR
To further compare the contents of volatiles and assess the correlation among the different drying methods, a heatmap was obtained after hierarchical cluster analysis (HCA) of the volatiles of the BR as shown in (Figure 6). The content value of each volatile was normalized by log2 transformation. The analysis was unsupervised. The Euclidean distance was used in the cluster analysis to display similarity among sample groups. The shorter the Euclidean distance, the greater similarities of the samples are. Similar squares were jointed together as they formed into cluster. Two major clusters were produced. The VFD group was in one cluster while all the other groups were in another cluster, suggesting the significant differences of VFD from others. Some of the key distinctly aroma volatile compounds were propanal, 2-pentylfuran, 1-(2-furanyl)-ethanone, nonanal, pentanal, and furfural, which were abundant in the HD, VD, and SD group, and less in the VFD group. HD50, HD60, and HD80 were in the same subclass while VD60, VD70, and VD80 were in other subclass. The effects of these temperatures were similar from HCA. SD was close to HD50 and VD50.

HCA of Volatile Compounds in BR
To further compare the contents of volatiles and assess the correlation among the different drying methods, a heatmap was obtained after hierarchical cluster analysis (HCA) of the volatiles of the BR as shown in (Figure 6). The content value of each volatile was normalized by log2 transformation. The analysis was unsupervised. The Euclidean distance was used in the cluster analysis to display similarity among sample groups. The shorter the Euclidean distance, the greater similarities of the samples are. Similar squares were jointed together as they formed into cluster. Two major clusters were produced. The VFD group was in one cluster while all the other groups were in another cluster, suggesting the significant differences of VFD from others. Some of the key distinctly aroma volatile compounds were propanal, 2-pentylfuran, 1-(2-furanyl)-ethanone, nonanal, pentanal, and furfural, which were abundant in the HD, VD, and SD group, and less in the VFD group. HD50, HD60, and HD80 were in the same subclass while VD60, VD70, and VD80 were in other subclass. The effects of these temperatures were similar from HCA. SD was close to HD50 and VD50.

Conclusions
In the current study, HS-GC-MS was used to analyze the volatile components in BR. Different drying methods including HD and VD at different temperatures, ND, SD, and VFD were investigated for their influence on the volatile compounds of BR by using the HS-GC-MS coupled with multivariate statistical analysis. The main volatile components were aldehydes, ketones, heterocycles, terpenes, aromatics, and so on. Different drying methods exerted different influences on types and the content of volatile compounds of BR. All the OPLS-DA models showed distinct separations, good interpretation, and predictability between the different groups. Results revealed that the most remarkable representative differential volatiles were dimethyl sulfide, 2-ethylfuran, furfural, and aldehydes. HCA suggested that the VFD was the most different drying procedure. The traditional drying methods (HD and SD) could be used to produce flavor substances as oxidation, Maillard reactions, and degradation reactions may take place during the drying process. The results obtained in this work provide an overall profile for understanding the volatile compounds in BR and also provide a guide for the drying, storing, and processing of BR.

Conclusions
In the current study, HS-GC-MS was used to analyze the volatile components in BR. Different drying methods including HD and VD at different temperatures, ND, SD, and VFD were investigated for their influence on the volatile compounds of BR by using the HS-GC-MS coupled with multivariate statistical analysis. The main volatile components were aldehydes, ketones, heterocycles, terpenes, aromatics, and so on. Different drying methods exerted different influences on types and the content of volatile compounds of BR. All the OPLS-DA models showed distinct separations, good interpretation, and predictability between the different groups. Results revealed that the most remarkable representative differential volatiles were dimethyl sulfide, 2-ethylfuran, furfural, and aldehydes. HCA suggested that the VFD was the most different drying procedure. The traditional drying methods (HD and SD) could be used to produce flavor substances as oxidation, Maillard reactions, and degradation reactions may take place during the drying process. The results obtained in this work provide an overall profile for understanding the volatile compounds in BR and also provide a guide for the drying, storing, and processing of BR.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/ 10.3390/foods10040868/s1, Table S1: The main 31 volatile components identified based on VIP > 1 and p-value < 0.05 between sun light drying(SD) and natural drying (ND), Table S2: The main 29 volatile components identified based on VIP > 1 and p-value < 0.05 between vacuum freeze drying (VFD) and natural drying (ND), Table S3: The main 27 volatile components identified based on VIP > 1 and p-value < 0.05 between hot air drying at 50 • C (HD50) and natural drying (ND), Table S4: The main 21 volatile components identified based on VIP > 1 and p-value < 0.05 between hot air drying at 60 • C (HD60) and natural drying (ND), Table S5: The main 25 volatile components identified based on VIP > 1 and p-value < 0.05 between hot air drying at 70 • C (HD70) and natural drying (ND), Table  S6: The main 30 volatile components identified based on VIP > 1 and p-value < 0.05 between hot air drying at 80 • C (HD80) and natural drying (ND), Table S7: The main 32 volatile components identified based on VIP > 1 and p-value < 0.05 between vacuum drying at 50 • C (VD50) and natural drying (ND), Table S8: The main 21 volatile components identified based on VIP > 1 and p-value < 0.05 between vacuum drying at 60 • C (VD60) and natural drying (ND), Table S9: The main 15 volatile components identified based on VIP > 1 and p-value < 0.05 between vacuum drying at 70 • C (VD70) and natural drying (ND), Table S10: The main 24 volatile components identified based on VIP > 1 and p-value < 0.05 between vacuum drying at 80 • C (VD80) and natural drying (ND). Figure  S1: The loading plots: (A) between natural drying (ND) and sun light drying (SD), (B) between ND and vacuum freeze drying (VFD); Figure S2: The loading plots for comparisons between the different temperatures of hot air drying (HD) and natural drying (ND): A for HD50 and ND, B for HD60 and ND, C for HD70 and ND, D for HD80 and ND; Figure S3: The loading plots for comparisons between the different temperatures of vacuum drying (VD) and natural drying (ND): A for VD50 and ND, B for VD60 and ND, C for VD70 and ND, D for VD80 and ND.