A Comparative Analysis of the Flavor, Aroma, and Compositions in Three Levels of Stir-Frying Cyperus rotundus Using E-Tongue, E-Nose, and HS-GC-MS

Cyperus rotundus (CR) is the dry rhizome of the Cyperaceae plant Cyperus , which is widely used in food processing, such as curry, pickling spices


Introduction
Cyperus rotundus (CR, known as Xiang Fu in Chinese), the dried rhizome of Cyperus rotundus L., is highly valued as both food and medicine [1]. In clinical practice, it is frequently prescribed for abdominal distention, irregular menstruation caused by liver depression, blood defciency and qi stagnation, irregular menstruation, chest, and abdominal pain [2]. In addition, it is a raw material for curry, salted spices, Greek dishes, and baked goods [3].
Stir-frying is one of the CR processing methods that have been used in China for more than 1340 years [4]. Troughout history, there have been primarily three levels of stir-frying: "slightly stir-frying," "stir-frying yellow," and "stir-frying black" [5][6][7]. However, relatively few studies have been conducted on stir-fried foods. Until now, the diferences between them have been unknown, preventing the full exploitation and application of sf-CR in food and TCM prescriptions. Tus, new research into three distinct levels of sf-CR is required.
Research on the processing mechanism is the foundation of food and traditional Chinese medicine (TCM) processing, as well as the key to TCM's internationalization and modernization. CR contains numerous chemical components that act on various pharmacological targets and regulate numerous biological mechanisms [8]. Consequently, only one or a few of the constituents provide a comprehensive explanation of their processing mechanism. Chemical component analysis has been increasingly applied to identify food and TCM [9]. Metabolomics is among the emerging "-omics" technologies [10]. It is consistent with the overall theory of Chinese medicine and may provide a novel approach to revealing the processing mechanism [11]. To date, numerous experimental studies on volatile components (VOCs) and the pharmacological efects of CR have been published. Accordingly, VOCs are believed to be the primary active components of CR. It is established that volatile components are one of the primary active components in CR [12,13]. Terefore, comparing VOCs for the three stages of sf-CR stands highly relevant. Several extraction techniques (carbon dioxide, supercritical fuid extraction, and hydro-distillation) have been reported [14][15][16] as essential steps in the analysis of volatile organic compounds (VOCs). However, the majority of these methods are time-and resource-consuming. In this regard, gas chromatography-mass spectrometry (GC-MS) comprises a direct and rapid approach to the identifcation of VOCs. In addition, headspace (HS) sampling has been demonstrated to be a rapid, simple, and convenient method for VOC analysis. HS-GC-MS combines the advantages of HS and GC-MS in terms of speed, convenience, and sensitivity and has, therefore, emerged as a potent instrument for analyzing VOCs.
Previous research in long-term medical practice has established a connection between TCM favor and aroma and its pharmacological efects [17]. According to modern medicine, the diference between the favor and aroma of TCM is a direct result of its chemical composition [18]. Terefore, evaluating the intrinsic properties of decoction pieces via favor can provide new perspectives and references for the prescription and clinical application of Chinese medical decoction [19]. Te evaluation of "favor and aroma" is gradually shifting from artifcial subjective feeling to digital and visual evaluation as a result of the continuous development of modern analysis technologies. Te advancement of bionic technologies, such as the electronic tongue and electronic nose, not only provides a tool for defning the favor and aroma of traditional Chinese medicine but also introduces new research methodologies to explain the mechanism of processing decoction pieces [20][21][22]. Currently, E-tongue and E-nose technologies are utilized extensively in the food, TCM, agricultural detection, and environmental monitoring industries [23][24][25][26][27]. However, no report exists to date on the use of the E-tongue and E-nose systems to evaluate the favor of CR.
Terefore, in the present study, we developed a method that efciently combines plant metabolomics, network pharmacology, and sensory bionic technology to investigate the changes in CR before and after vinegar processing. Using HS-GC-MS and a data analysis strategy based on metabolomics, we then screened, identifed, and compared the major volatile compositional diferences between RCR and VCR. Finally, the diferences in aroma between the three types of sf-CR were explained using Spearman's rank correlation between aroma and volatile compounds. Te fndings are expected to clarify the changes in favor, aroma, and volatile components of CR during stir-frying, as well as the process mechanism, in order to lay a foundation for its application in foods and formulations.

Reagents and Materials.
Tartaric acid (analytical grade) was purchased from Tianjin Oberkai Chemical Co., Ltd., whereas potassium chloride (analytical grade) was procured from Tianjin Best Chemical Co., Ltd. Deionized water used in the study was obtained used a ULUPURE series ultrapure water machine (Model: UPR-II-20L). In addition, positively and negatively charged membrane washing solutions were purchased from Beijing Ying Sheng Heng Tai Technology Co., while the internal solution (211119) was procured from Beijing Ying Sheng Heng Tai Technology Co. Methanol of HPLC grade was attained from Fisher Scientifc (Fair Lawn, NJ, USA); cyperotundone (DST201120-156, 98%) and α-cyperone (DST200819-022).

Sample Collection and Processing. Ten batches of raw
Cyperus rotundus were collected from the Anguo medicinal materials market in Hebei province, which were authenticated by the Chief Chinese pharmacist Jiping Duan (Hebei Institute of Drug and Medical Device Control). Te voucher specimens were then deposited at the Hebei Traditional Chinese Medicine Formula Granule Engineering and Technology Innovate Center, Shijiazhuang, China. Tese 10 batches of raw-RC were subsequently utilized to prepare 30 batches of sf-CR. Te batches were categorized as follows: (a) slightly stir-fried (LY1-LY10)-raw-RC granules were slightly stir-fried over a mild fre for 5 min until the surface turned light yellow in color; (b) stir-frying yellow (Y1-Y10)-raw-RC granules were stir-fried over a mild fre for 15 min until the surface turned yellow in color; (c) stirfrying black (B1-B10)-raw-RC granules were stir-fried over a mild fre for 30 min until the surface turned dark brown in color. All samples were dried at room temperature and crushed to obtain a particle size between 50 (355 ± 13 μm) and 80 meshes (180 ± 7.6 μm).

E-Tongue System and Sampling
Procedure. Te E-tongue (SA402B, INSENT, Japan) consisted of a sensor array, special beakers, signal acquisition, and data analysis software packages and was used to detect the favor. Te sensor array consisted of 6 sensors (AAE, CTO, CAO, COO, AE1, and GL1) that were sensitive to the umami, salty, sour, bitter, astringent, and sweet favors, respectively. Te sensor unit measured the reference solution for 30 s, and at the conclusion of the measurement, the electrical potential was recorded. Te detector unit then measured the sample solution in a similar manner. After measuring the sample solution, the sensor unit was briefy rinsed in the stock solution, and the reference solution is remeasured for 30 s using the detector unit. Te sensor unit was fnally washed with the wash solution and fushed with the reference solution for subsequent measurements.
1.0 g of the sample was collected and placed in a conical vial (100 mL) equipped with a stopper. Ten, 100 mL of pure water was added with sonicating for 30 minutes. Te resultant mixture was then added to a 100 mL centrifuge tube and centrifuged for 10 min at 4000 rpm•min −1 . A special beaker was then used to test ∼35 mL of the as-obtained supernatant. Finally, four replicates of the sample were analyzed, and the raw data from the last three replicates were selected for analysis [28].
1.0 g of the samples were measured and placed in a 100 ml glass beaker. Te samples were then sealed using two layers of sealing flm and allowed to rest at room temperature for 30 min to ensure that the sample aroma was acceptable and that its test procedure was equilibrated under the following conditions: fush duration-120 s; sample interval-1 s; zero-point trim time-10 s; presampling time-5 s; measurement time-120 s; and gas fow rate--600 mL•min −1 [29,30]. Each sample was analyzed in triplicate.

HS-GC-MS Conditions.
Te headspace injection conditions were as follows: a 20 mL headspace inlet vial containing 0.5 g of sf-CR powder was sealed and incubated at 120°C for 30 min.
Te gas chromatography conditions are outlined as follows: volatile components of sf-CR samples were analyzed using a GC 2030 gas chromatograph (Shimadzu, Japan). Te volatile components were isolated on an SH-I-5SiI MS (30 m × 0.25 mm × 0.25 μm, Shimadzu, Japan). Te temperature program began at 50°C; it was then raised to 70°C at a rate of 2°C/min. Subsequently, it was raised to 120°C at a rate of 10°C/min, following which it was raised to 210°C at a rate of 5°C/min and maintained for an additional 2 min. Te carrier gas comprised helium (purity 99.99%) with a constant fow rate of 1.5 mL/min. Te injection volume was 1 μL with 1 : 20 splits.
Te mass spectrometry conditions were as follows: MS detection was carried out on a TQ8050 NX mass spectrometer (Shimadzu, Japan). Te ion source temperature was 230°C, while the interface temperature was 250°C. In addition, a mass scan range of 50-600 m/z was utilized, while the solvent delay was maintained at 2.5 min.

Quality Control Measurement.
Quality control (QC) samples were obtained by mixing 0.5 g of each sample in order to validate the repeatability and stability of the assay system. A total of eight consecutive QC samples were tested to evaluate the stability and reliability of the system and methods. During the formal testing process, every ffth sample is interspersed with a QC sample to assess repeatability and precision.
2.6. Data Analysis. Te statistical analysis was conducted using the Statistical Package for Social Sciences 19.0 (SPSS Inc., Chicago, IL, USA). SIMCA-P14.1 (Umetrics AB, Umea, Sweden), GraphPad prism (8.0, San Diego, California, USA), and ChiPlot Online platform (https://www.chiplot.online/) were utilized for the statistical analyses (PCA, OPLS-DA, HCA, and group error bar plot) [31]. For peak extraction, MS-DIAL 4.90 was used [32]. VIP (OPLS-DA) was obtained using SIMCA-P14, while P value (T-test) and FC (fold change) were obtained using Excel [33]. Finally, VIP > 1, P < 0.05, and FC ≥ 2 or FC ≤ 0.5 as the threshold was used to obtain the diferential compounds [34][35][36]. Te total differential compounds are the union of three diferential compounds. Te identifcation of diferential compounds was performed using NIST 20.0 database and retention indexes (RI) [37]. First, we matched the experimental data with the database, including RI, secondary mass spectrometry fragments, and similarity parameters. Te obtained mass spectrum is compared with the known compounds in the database to determine the name, molecular formula, and other information of the compounds. Ten, the obtained ingredient information is verifed with the information reported in the literature.

E-Tongue Analysis.
Te output value generated by the reference solution was 0. Except for sourness and saltiness, every other indicator had a favorless value of 0. Furthermore, the favorless points for sourness and saltiness were −13 and −6, respectively, because the reference solution was composed of potassium chloride and tartaric acid, which contain a small quantity of acid and salt [38,39]. Figure 1(a) demonstrates that the saltiness and sourness of the samples fall below the criterion for favorlessness. Additionally, astringency, aftertaste, and richness had values that were close to the favorless point, whereas bitterness, umami, and sweetness had values that were signifcantly greater than the favorless point (Table S1). Subsequently, we examined various efective indicators after removing the saltiness and sourness values.
Based on the principal component analysis (Figure 1(b)), the frst principal component (explaining 58.3% of the variance) distinguished light stir-frying from yellow and black, whereas the second principal component (explaining 20.7% of the variance) distinguished stir-frying yellow from black. Correspondingly, the results demonstrated signifcant diferences in favor between groups, with the greatest difference between the light stir-frying group and the other two groups, in addition to a relatively small diference between the yellow and black groups. Hierarchical clustering analysis (HCA) is a statistical technique for identifying relatively homogeneous sample groups based on particular  characteristics [40]. Figure 1(c) depicts the favor HCA for each sample, wherein the horizontal axis represents the sample, and the vertical axis represents the distance. Correspondingly, sf-CR samples were divided into two clusters: one for yellow and black stir-frying and the other for light stir-frying. As observed, the fndings were consistent with the principal component analyses (PCA) results mentioned previously.
A one-way ANOVA was conducted to gain additional insight into the favor that contributed to the aforementioned diferences (Figure 1(d)). Except for aftertaste-B (bitterness aftertaste), a signifcant diference in the sf-CR for six of the seven favors was observed. Te response values for sweetness, bitterness, astringency, umami, and richness were signifcantly greater in the slightly sf-CR than in the yellow and black ones (P < 0.05). Additionally, the stir-frying yellow and black groups difered signifcantly only in aftertaste-A (astringent aftertaste), umami, and richness (P < 0.05), whereas there was no signifcant diference in sweetness, bitterness, and aftertaste-B. Overall, the favor of CR decreased as stir-frying grades increased.

E-Nose Analysis.
Te average response values of E-nose sensors for each sample are provided in Table S2. Compared to other sensors, the W5S, W2W, and W1W exhibited a signifcantly larger magnitude of response to each sample. Correspondingly, the diference between the three sf-CR groups was investigated using univariable and multivariable analyses.
Te distribution of the PCA score plot (Figure 2(a)) revealed that the stir-frying lightly, yellow, and black groups exhibited distinct clustering. Furthermore, the yellow stir-fry group was much closer to the light stir-fry group than the black group, implying that the more intense the stir-frying, the greater the aroma change. In addition, the variance contribution rate of the frst principal component PC1 was 47.1%, and the variance contribution rate of the second principal component PC2 was 19.1%; cumulatively, they accounted for 66.2% of the information in the original data set cumulatively. When combined with the loading plot (Figure 2(b)), sensors W1S and W6S contributed the most to the frst principal component, followed by sensors W2W and W1W. Additionally, sensor W2S made the greatest contribution to the second principal component. Similar comparisons were also performed using hierarchical clustering analyses with very similar outcomes (Figure 2(c)). Correspondingly, the 30 batch samples were divided into 3 major clusters: slightly stir-fried, yellow, and black, respectively. Both principal component analysis and hierarchical clustering analyses revealed statistically signifcant diferences between the three sample groups.
A one-way ANOVA was conducted to learn more about the aroma responsible for the aforementioned diferences; the results are depicted in Figure 2(d). Six of the ten sensors, excluding W1C, W3C, W5C, and W3S, exhibited a signifcant variation in sf-CR. Tis may be the primary reason for the diferences between the three sf-CR groups. As shown in Table 1, the content of nitrogen oxides, hydrogen, metal, terpene, and aromatic compounds difered signifcantly between the three groups of sf-CR. In addition, the yellow and black stir-fry samples had a stronger aroma than the light stir-fry samples (P < 0.05); in particular, the diference in aroma associated with W6S, W1W, and W2W was signifcantly stronger (P < 0.01). Te combination of W5S, W6S, W1S, W1W, and W2W was primarily responsible for the negligible diferences in aroma between stir-fried yellow and black CR samples (P < 0.05). Moreover, the W1W sensor aroma of black stir-fry samples were less than that of yellow samples, while the aroma of all other samples increased, with the W1S and W2S enhancement tendencies being more signifcant (P < 0.01). In conclusion, as the stirfrying grade increased, the aroma of CR steadily increased.
Modern chemistry recognizes favor as one of the predominant sensory properties. Multiple compounds interact intricately to produce favor and aroma. Since favor and aroma are chemosensory parameters, the senses are just as important as the chemical components [41]. Te E-tongue and E-nose closely mimic the perception and evaluation of favor and aroma by the human tongue and nose and objectively evaluate them [42,43].

Screening of Diferential Compounds.
As depicted in the PCA plot (Figure 3(a)), QC samples clustered together, indicating that the instrument was stable, and the method was repeatable. PCA analysis revealed that 30 batches of SF-CR samples were separated into three groups: lightly stirfried, yellow, and black, indicating that VOCs varied among the three groups to identify the diferential VOCs responsible for the diferences in sf-CR. We frst performed a preliminary screening of the potential diferential VOCs based on P < 0.05 and FC ≥ 2 or ≤0.05, and we visualized the screening results using volcano plots. A total of 38 potential diferential compounds (3 upregulated and 35 downregulated) were identifed between the yellow stir-fried and lightly stir-fried CR samples (Figure 3(b)); in addition, 94 potential diferential compounds (48 upregulated and 56 downregulated) were identifed between the lightly stir-fried lightly and black stir-fried CR samples (Figure 3(c)), and 56 potential diferential compounds (20 upregulated and 36 downregulated) were identifed between the stir-fried yellow and stir-fried black CR samples (Figure 3(d)).
OPLS-DA is a method of multivariate statistical analysis with supervised pattern recognition that can efectively flter out diferential VOCs by removing irrelevant infuences from the search [44]. As depicted by the OPLS-DA scatter plot (Figures 4(a), 4(d), and 4(g)), pairs of samples from the stir-frying lightly, yellow, and black groups were more easily distinguished more clearly in pairs, as shown in the OPLS-DA scatter plot (Figures 4(a), 4(d), and 4(g)).  LY5  LY6  LY7  LY8  LY9  LY10  LY2  LY3  Y1  Y4  Y2  B10  Y5  Y6  Y9  Y10  B3  B9  B5  B7  B6  B4  B2  B8  Y8 Y7 B1 Y3     Journal of Food Biochemistry deemed to be the primary VOCs responsible for the differences between the three levels of sf-CR. As a result, 20, 26, and 15 VOCs were fltered out from the light stir-fry vs. yellow group, light stir-fry vs. black group, and yellow vs. black group. When compared with the stir-frying lightly group, 14 VOCs in the yellow group decreased while 6 VOCs increased; furthermore, in the black group, 19 decreased and 7 increased. In addition, when compared to the yellow stirfry group, 12 VOCs in the group black decreased, while 3 VOCs increased. Subsequently, after removing the repeated VOCs from the three groups, a total of 33 unique VOCs were identifed.

Identifcation and Analysis of Diferential Compounds.
On the basis of the NIST 20.0 database, GC-MS postrun analysis was utilized to determine 33 key diferential compounds. 21 sesquiterpenes, 8 monoterpene hydrocarbons, and 4 pyrazines were identifed. Te information included the compound name, RI, formula, category, molecular weight, as presented in Table 2 (the GC-MS raw data of diferential compounds are shown in Table S3). Using a heat map ( Figure 5(a)), we analyzed the distribution of the various compounds present in sf-CR samples of varying concentrations in order to determine their presence. Ten samples processed using the same method at three diferent levels of stir-frying were clustered into a single group, indicating excellent homogeneity among biological duplicates and high data reliability. In the various sf-CR samples, the accumulation of these compounds exhibited a distinct phenotypic variation with regard to their abundance.
Based on their accumulation patterns, these compounds could be clearly categorized into four major clusters. Te relative concentration of components in clusters 1 and 4 increased with increasing degree of frying in all samples. On the other hand, the relative concentration of components in clusters 2 and 3 decreased as the degree of frying increased. Tus, it is possible that as stir-frying is intensifed, the concentration of certain terpenoids decreases while the concentration of pyrazine rises. Notably, these pyrazine compounds were not reported in the CR study; however, they may be the new components of CR after stir-frying.
It is worth noting that aromatic herbs produce pyrazine compounds when fried. Apparently, it results from the Maillard reaction between amino acids and carbohydrates [45]. During roasting, the Maillard reaction occurs at relatively high temperatures and low levels of moisture. In the Maillard reaction, moisture is a crucial factor that determines the direction of the reaction and the quantity of volatile compounds produced via this pathway [46]. Tis is one reason that the aroma profles of the roasted and boiled samples were distinct [47,48]. With the extension of stir frying time, the moisture content in CR decreased, which promoted the occurrence of Maillard reaction. Consequently, as stir-frying intensifes, pyrazine levels rise. Terpenoids are highly volatile; however, in stir-frying processing, the temperature is elevated, and some components are susceptible to oxidative decomposition, rearrangement, isomerization, and other reactions, which will result in a reduction of terpenoids. Intriguingly, there is also an increase in the content of a small number of terpenoids, which we hypothesize may be the result of transformations between complex constituents that require further investigation.

Correlation Analysis of Diferential VOCs and E-Nose
Sensors. In addition to the above, we assessed the relationships between 33 key diferential VOCs and the aroma of sf-CR using Spearman's rank correlation ( Figure 5(b)). Te results revealed that the diferential compounds were most strongly correlated with sensors W3C, W5C, and W3S. Tese three sensors are activated by ammonia, aromatic compounds, and aromatic compounds, alkenes, respectively (Table 1). Of these, 2-methyl pyrazine (P1), 1,4-dimethylpyrazole (P2), 2,5-dimethylpyrazine (P3), 2-methoxy pyrazine (P4), D-limonene(P8), camphene (P9), transverbenol (P10), and pinocarvone (P11) had positive correlations with W3C and W5C response values and negative correlations with W3S. Other components, excluding pcymene and copaene, exhibited a negative correlation with the response values of W3C and W5C but a positive correlation with W3S. Tese were the principal volatile favor compounds present in sf-CR. As observed, these fndings were consistent with the results of the diferential  Journal of Food Biochemistry component heat map. As the intensity of stir-frying increased, the proportion of pyrazines and a small amount of terpenoids (P8, P9, P10, and P11) increased, and the aroma intensifed accordingly. Based on correlation analysis, we found the main compounds that cause aroma diferences. Pyrazines are nitrogen-containing heterocyclics characterized by their potency [49]. Due to their low vapor pressure, pyrazines readily evaporate [46]. Numerous of these substances have intense aromas and extremely low aroma threshold values [46]. Consequently, these compounds contribute to a variety of favors. Te results of this study revealed a direct correlation between sf-CR aroma and pyrazine content. In addition, it has been discovered that some pyrazines have pharmacological efects, including antithrombogenic, diuretic, and hepatoprotective [50][51][52]. In addition, numerous favor volatiles have been derived from terpenoids unrelated to nutrients [53]. However, many of these terpenoids are also known to possess antimicrobial properties. Tese fndings may, therefore, provide a foundation for future research on the application of sf-CR in food and medicine at varying concentrations.

Conclusion
In this study, an integrated method combining E-tongue, Enose, and HS-GC-MS was developed for the frst time in order to investigate the distinctions between three distinct levels of sf-CR. As the intensity of stir-frying increased, the favor of sf-CR diminished while its aroma increased. Accordingly, the three types of sf-CR were screened for and identifed with 33 distinct compounds, including 29 terpenoids and 4 pyrazines. Te proportion of terpene components decreased as stir-frying intensity increased, while the proportion of pyrazine components increased. Additionally, Spearman's rank correlation allowed us to identify 2-methylpyrazine, 1,4-dimethylpyrazole, 2,5-dimethylpyrazine, 2-methoxypyrazine, D-limonene, camphene, transverbenol, and ponocarvone as the primary components causing aroma diferences. Tus, the current study   established a method for identifying the diferences in favor, aroma, and volatile components of the three levels of sf-CR, as well as screening the major components responsible for aroma diferences. Tis provides a reference for the use of sf-CR in food processing and medicine, as well as new ideas and methodologies for the investigation of other foods and pharmaceuticals.