Characterization of the key volatile organic components of different parts of fresh and dried perilla frutescens based on headspace-gas chromatography-ion mobility spectrometry and headspace solid phase microextraction-gas chromatography-mass spectrometry

3.1.1 Differences of VOCs between fresh and dried P. frutescens by HS-GC-IMS

HS-GC-IMS analysis was the latest non-targeted and rapid analytical technique which had great potential for detecting small-molecule(Chen et al., 2021). For exploring the differences of small-molecule VOCs among fresh and dried P. frutescens, the small-molecule VOCs were detected by HS-GC-IMS. In the present study, 60 small-molecule VOCs were identified according to RI and Dt in all P. frutescens samples by HS-GC-IMS (Supplementary Information, Table A.1 and Fig.S1). Due to the adduct formation between partial molecules and reactant ions, a single VOCs with high concentration or higher proton affinity than that of water could form multiple signals presenting as dimer or polymer(Zhang et al., 2021). Therefore, the identified components included both dimers and monomers. However, the differences among fresh and dried P. frutescens could not be quickly identified by table. Therefore, the fingerprints were established between fresh samples and dried samples in the same part (Fig. 1A&B&C). The color represented the peak intensity. The peak intensity increased from blue to white and then to red. The compositional differences were evident between fresh and dried samples in fingerprints. 22 small-molecule VOCs (heptan-2-ol, vanillin, benzyl alcohol, etc.) only were identified in fresh stem (Fig. 1A, a), Conversely, 10 small-molecule VOCs (e.g., 4-methylguaiacol, coumarin, 4-ethylphenol, etc.) were recognized only in dried stems (Fig. 1A, b). Compounds “Unknown” were unidentified compound. After comparing fresh and dried P. frutescens stems, it was found that fresh stems had many special components that dried stems do not have. For example, benzothiazole of fresh stems had anti-cancer, anti-inflammatory, analgesic antidiabetic, etc.(Das et al., 2022), which suggested that future research could develop related products for fresh P. frutescens stems. 18 small-molecule VOCs (p-cymen-7-ol, nerol, eugenol, etc.) were only identified in fresh leaves (Fig. 1B, c). Nerol and eugenol with antibacterial and antioxidant properties had long been used in food and cosmetology(Wang et al., 2019, Ulanowska and Olas 2021). 12 small-molecule VOCs were only identified in dried leaves, such as neryl acetate, ethyl 3-phenylpropanoate, and so on (Fig. 2B, d). Vanilla was detected both in fresh and dried leaves, which was the world’s most popular flavor in food, cosmetics, and medicine(Arya et al., 2021). 5 small-molecule VOCs (α-terpineol, 2-acetylthiazole, etc.) were identified exclusively in dried seeds (Fig. 1C, e), while geraniol and phenyl ethyl alcohol were just recognized in fresh seeds (Fig. 1C, f). Geraniol were regarded as good natural preservatives on account of antibacterial and antioxidant(Lin et al., 2021). After drying, the components existing in the fresh samples disappeared in the dried sample, and many new components were produced in the dried samples (Fig. 1A&B&C), especially in the stems and leaves samples, which demonstrated those compounds were sensitive to drying process and were decomposed or degraded. Besides, these components may contribute to the corresponding aroma between fresh and dried P. frutescens. The signal strength represented the concentration of flavor molecules (Fig. 1A&B&C) of HS-GC-IMS, the white color indicated the lower concentration while the red color box showed higher and the darker the color gets, the higher the concentration of fingerprint plots(Chen et al., 2021). Compared to dried samples (seeds, leaves and stems), fresh samples (seeds, leaves and stems) seemed to have a stronger signal intensity. The changes of signal intensities of small-molecule VOCs were due to the chemical reactions (Maillard reaction, interaction between different molecules and degradation of macromolecules) during drying(Guo et al., 2018). Fresh P. frutescens was more suitable as flavorings or vegetables to use because it had a strong flavor concentration. Besides, the small-molecule VOCs that had been identified were classified into six ingredient categories (Fig. 1D). Alcohols (11 components) and others (7 components) were most found in fresh P. frutescens leaves. Esters (9 components) were most found in dried P. frutescens leaves. Ketones (6 components) mainly focused on dried P. frutescens seeds. Aldehydes (3 components) were relatively abundant in fresh P. frutescens seeds. Thus, the aroma of P. frutescens were more derived from alcohols and Esters. In the current study, most of the small molecules contained in P. frutescens had strong oxidative and antibacterial bioactivities, making it a potential ingredient for functional food development. Unfortunately, there were still many small-molecule VOCs of P. frutescens that had not yet been identified. Therefore, the IMS database needed to be further supplemented in future researches.

Close

Fig. 1

Fingerprints and radar chart of the fresh and dried P. frutescens from HS-GC-IMS. (Each row stands for a selected signal peak in one sample and each column stands for a signal peak of a compound in a different sample. Fst: fresh stems (a), FL: fresh leaves (b), FSe: fresh seeds (c), DSt: dried stems (d), DL: dried leaves (e), DSe: dried seeds (f). (A) stems, (B) leaves, (C) seeds. (D) Component classification identified by HS-GC-IMS.

Export to PPT

3.1.2 Differences of VOCs between fresh and dried P. frutescens by HS-SPME-GC–MS

HS-SPME-GC–MS was used widely and effectively for analyzing VOCs. In order to explore comprehensively VOCs between fresh and dried P. frutescens, HS-SPME-GC–MS was performed. A total of 115 VOCs were identified by HS-SPME-GC–MS (Supplementary Information, Table A.2). Seven compounds had passed standard validation (Fig. 2A, a), and the remaining components had been identified through literatures, database, and RI. The components detected by HS-SPME-GC–MS were more than HS-GC-IMS. The total ion chromatograms (TIC) of all P. frutescens samples were shown in Supplementary Information (Supplementary Information, Fig. S2). Elsholtzia ketone, cuminaldehyde, perilla ketone, caryophyllene, and linalool with relatively high content(Fig. 2A, b) possessed pharmacological activities for cardiovascular, respiratory systems and tumor cells(Ahmed 2018, Chen et al., 2020). After natural shade drying, these VOCs displayed change to some extent. Compared to the fresh P. frutescens stems, a significant increase in the content of elsholtzia ketone, perilla ketone, caryophyllene, and linalool was found in the dried P. frutescens stems. Compared with the fresh P. frutescens leaves, the contents of perilla ketone, caryophyllene, and linalool were increased in all samples of dried P. frutescens leaves, but elsholtzia ketone and cuminaldehyde were decreased in dried P. frutescens leaves excluding sample 4 which had a slight increase. Elsholtzia ketone and cuminaldehyde were decreased in sample 5 and increased in the remaining dried P. frutescens seeds compared with the fresh P. frutescens seeds, whereas caryophyllene was decreased except for sample 5. Two other compounds (perilla ketone and linalool) were both increased and decreased between different batches of dried P. frutescens seeds. Interestingly, elsholtzia ketone and cuminaldehyde content seemed to exhibit synergistic effects after shade drying whether leaves and stems or seeds of P. frutescens. The differences may be a factor of dried P. frutescens as medicine. Some VOCs such as (R)-l-perillaldehyde, α-farnesene and trans-2-hexenal were only detected in dried stems. Conversely, some VOCs such as (R)-cuparene, beta-elemene and citral were only identified in fresh stems. Cis-linalool oxide, (R)-l-perillaldehyde and α-cubebene were just detected from dried leaves while butylated hydroxytoluene, α-longipinene, and 2-methylbutyraldehyde were only tested from fresh leaves. 1-(6,6-dimethylbicyclo[3.1.0]hex-2-en-2-yl)-ethanone and ylangenol were only found in the dried seeds. Conversely, some ingredients (e.g. 3-carene, β-guaiene, egomaketone, etc.) were only detected in the fresh seeds. The VOCs formed after drying may be attributed to precursors in fresh samples being degraded, isomeric species or oxidation reactions during the drying process(Song et al., 2021). Furthermore, small-molecule VOCs detected by HS-GC-IMS may be involved in synthesis and conversion processes of the aforementioned compounds detected by HS-SPME-GC–MS. For example, limonene could produce α-terpineol by biotransformation(Molina et al., 2019). Terpenes were relatively abundant in fresh and dried P. frutescens (Fig. 2D). Terpenes were a vital member of the VOCs safeguarding plants against pests, diseases, and weeds, and also offered a wide range of prospects in the pharmaceutical, food, cosmetic, and flavoring industries(Ninkuu et al., 2021).

Close

Fig. 2

The total ion chromatograms (TIC) (A) and radar chart (B) of the fresh and dried P. frutescens from HS-SPME-GC–MS. (A): The total ion chromatograms of mixed standard chemicals (a) and dried P. frutescens leaves samples (b). 1: 1-Octen-3-ol; 2: Linalool; 3: Perilla ketone; 4: Citral; 5: Caryophyllene; 6: Humulene, 7: Caryophyllene oxide. (B): Component classification identified by HS-GC-IMS HS-SPME-GC–MS detected. Fst: fresh stems, FL: fresh leaves, FSe: fresh seeds, DSt: dried stems, DL: dried leaves, DSe: dried seeds.

Export to PPT

3.2.1 Differences of VOCs in three parts of fresh or dried P. frutescens by HS-GC-IMS

Dried parts such as leaves, stems and seed were regarded as different drugs from ancient times to the present, but the underlying reason remains unclear. The fingerprint profiles form by HS-GC-IMS (Fig. 3) indicated there was distinct variability among different parts of P. frutescens. Among the different parts of fresh P. frutescens, the part with the largest amount and content of components was the fresh leaves (Fig. 3A, b), and the least was the fresh seeds (Fig. 3A, c). The flavor concentration of P. frutescens leaves was significantly higher, making them one of the favorite vegetables. The researches on different parts of fresh P. frutescens were very meaningful and could promote the development from multiple perspectives in the food field. Among the different parts of dried P. frutescens, the part with the largest number of components was the dried seeds (Fig. 3B, f), and the least was the dried stems (Fig. 3B, d). The differences in the amount and content of components in different parts of dried P. frutescens may be one of the reasons why different parts (seeds, leaves and stems) of P. frutescens as different drugs from the results (Fig. 3B). After the dataset was reduced dimensionality, principal component analysis and hierarchical cluster analysis (HCA: Fig. 4A&B) and (PCA: Fig. 4C&D) were carried out for chemometric analysis. Linkage method of HCA was average-link. Correlation was selected as type of distance. As displayed by the results of HCA (Fig. 4A&B), all fresh samples could be divided into three groups (I, II and IV) which correspond to the seeds, leaves and stems except for two of the samples (stem 5 and leaves 4) was misassigned (Fig. 4A). Meanwhile, other dried samples were clustered into three major clusters (I, II and IV) according to three medicinal parts (seeds, leaves and stems) although three samples of stem 4, leaves 4 and leaves 5 were not clustered into corresponding clusters (Fig. 4B). The result of PCA (Fig. 5C: PC1:50.9%, PC2:20.3%, PC3:9.6%; Fig. 4d: PC1:40 %, PC2:25.9%, PC3:16.5%) was similar to HCA, but HCA relatively achieved better results. Overall, fingerprint profiles, PCA and HCA indicated that VOCs of small molecules in P. frutescens from different parts (leaves, stems and seeds) showed apparent differences. At present, some small-molecule VOCs had been reported to contribute to anti-cancer, such as geraniol, benzothiazole, geranyl acetate, and so on(e Silva et al., 2022). These small-molecule VOCs differences may be associated with the treatment of different kinds of diseases. Thus, the small-molecule VOCs of P. frutescens in pharmacological studies deserved further investigation. Interestingly, although all samples came from the botanical gardens, there were certain differences between batches indicating that using HS-GC-IMS was able to detect subtle differences, which the instrument provided a reference for quality control of P. frutescens. Meanwhile, P. frutescens had genetic instability and easily mated with closely related species to form similar morphology, but there were obvious differences in chemical composition(Hu et al., 2010). Thus, it also suggested more samples were necessary to explore the cultivation conditions, harvesting time, and processing conditions of high-quality P. frutescens to guarantee stable quality.

Close

Fig. 3

Fingerprints among three different parts of P. frutescens from HS-GC-IMS. (Each row stands for a selected signal peak in one sample and each column stands for a signal peak of a compound in a different sample). Fst: fresh stems (a), FL: fresh leaves (b), FSe: fresh seeds (c), DSt: dried stems (d), DL: dried leaves (e), DSe: dried seeds (f). (A) Different parts of fresh P. frutescens, (B) Different parts of dried P. frutescens.

Export to PPT

Close

Fig. 4

(A)&(B): Dendrograms of different parts resulting from HCA. (C)&(D): Classification of different parts analyzed samples by PCA analysis. FSt: fresh stems; FSe: fresh seeds; FL: fresh leaves; DSt: dried stems; DSe: dried seeds; DL: dried leaves.

Export to PPT

Close

Fig. 5

OPLS-DA and RF of different parts of P. frutescens (blue: seeds, green: stems, red: leaves). FSt: fresh stems; FSe: fresh seeds; FL: fresh leaves; DSt: dried stems; DSe: dried seeds; DL: dried leaves. (A)&(B) Discriminatory classification map. (C)&(D) The variable importance for the projection (VIP) predictive of the VOCs (a&b were VIP value greater than 1). (E) significant features of different parts of the fresh. (F) significant features of different parts of the dried. Additionally, (1R,4R,5S): (1R,4R,5S)-1,8-dimethyl-4-(prop-1-en-2-yl)spiro[4.5]dec-7-ene. (1S,4S,4aS): (1S,4S,4aS)-1-isopropyl-4,7-dimethyl-1,2,3,4,4a,5-hexahydronaphthalene. (2R,3S,5S,6R): (2R,3S,5S,6R)-2,5-bis(4-Methoxyphenyl)-3,6-dimethyl-1,4-dioxane-rel- 1,1,4,7-Tetramethyl: 1,1,4,7-Tetramethyl-1a,2,3,4,6,7,7a,7b-octahydro-1H-cyclopropa[e]azulene.

Export to PPT

3.2.2 Differences of VOCs of three parts in fresh or dried P. frutescens by HS-SPME-GC–MS

Although there were no significant differences in the types of VOCs identified by HS-SPME-GC–MS, differences in content of components of different parts were observed (Supplementary Information, Fig.S2 and TableA2). Since there was genetic diversification and varied the content of components in VOCs, previous studies based on the main component content of essential oil divided into the following chemotypes: perillaketone (PK), perillaldehyde (PA), elsholtziaketone (EK), citral (C), phenylpropanoids (myristicin, dillapiole, elemicin) (PP), D-limonene and piperitone (DLP), β-caryophyllene myristicine (MT), perillene (PL) and piperitenone types (PT)(Kimura and Ito 2020). In this study, cuminaldehyde, elsholtzia ketone, perilla ketone and caryophyllene were mainly detected. The content of these components was ranked in the top five of the total VOCs in most of the samples either fresh or dried P. frutescens. Thus, there were two chemotypes identified, including PK-type and EK-type according to the main VOCs in the HS-SPME-GC–MS data. Besides, cuminaldehyde made up the largest content of component in some samples, which had been reported that according to a new classification of five essential oils based on cluster analysis of Euclidean distance, it was divided into fifth categories (EK chemotypes)(Wei et al., 2007). Therefore, the results of the current study supported with classification finding of Euclidean distance analysis and provided a scientific reference for EK chemotypes. Chemotypes of P. frutescens were related to the harvest period, sowing period and growing environment(Ghimire et al., 2017). Therefore, the result appeared variations of these VOCs between batches and had two chemotypes in the collection place.

Different parts of P. frutescens had been used in different medicine clinically. In order to further explore the different VOCs among different parts and screen out suitable volatile markers, the study combined orthogonal partial least squares discriminant analysis (OPLS-DA) and the Random Forest (RF) for analysis. The RF was a supervised method to handle high-dimensional data. The model displayed the best overall performance on volatile markers-selected (Chen et al., 2013). OPLS-DA was commonly used for visualizing high dimensional data and discriminant analysis of latent metabolites related to metabolic changes(Qi et al., 2020), which proved to be powerful tools in qualitative data analysis. Moreover, the variable importance of the projection (VIP) in OPLS-DA was formed by measuring variables with scores. VIP ≥ 1 was used to identify the variables important to the model and were generally considered to be important variables. The larger the VIP value, the more important the variable to the divergence was between different parts. Meanwhile, VIP ≥ 1 was considered to be a significant factor for samples classification(Galindo‐Prieto et al., 2014). In the study, two HS-SPME-GC–MS datasets (different parts of fresh P. frutescens and different parts of dried P. frutescens) were calculated by SIMCA 14.1 software based on the OPLS-DA model respectively (Fig. 5). All P. frutescens samples could be divided into three groups according to the different medicinal parts (seeds, leaves and stems) whether fresh (Fig. 5A) or dried P. frutescens (Fig. 5B), which indicated that differences existed among three parts. The explanatory and predictive abilities of the models were evaluated from R² and Q². And the closer R² and Q² were to 1, the better fitness of the model was (Wang et al., 2019). The R² and Q² of the two OPLS-DA models were greater than 0.9 (R² = 0.978, Q² = 0.915 and R² = 0.965, Q² = 0.92), which indicated excellent explanatory and predictive effects of the OPLS-DA model for the classification of different parts of P. frutescens. Based on the criteria of VIP ≥ 1, 56 (a) and 60 (b) important variables were selected in different parts of fresh and dried P. frutescens respectively from the VIP plot of OPLS-DA (Fig. 5C&D). To further explore favorable volatile markers from these essential variables, a RF algorithm was used to build RF model. The classification trees of RF model were set as 1000. During trees building, one-third of the samples were used as training data and the remaining samples were used as test data to get a judicial assessment of the out-of-bag (OOB) error. Theoretically, the cross-verification was conducted internally in the RF algorithm between the operation by the OOB dataset. The lower OOB error was a criterion of classifier quality(Bulgarevich et al., 2018). After several trees, the cumulative OOB error rates decreased to 0.0667 in different parts of fresh P. frutescens and to 0 in different parts of dried P. frutescens. Fig. 5E&F showed the first fifteen significant features in the ranking through the RF model. γ-muurolene, α-cubebene, △-cadinene, copaene, β-guaiene and germacrene D were more abundant in fresh seeds (Fig. 5E) with antibacterial and antioxidant, which were suitable for exploitation in various skin-care products(Bahadori et al., 2017, Falleh et al., 2020). (R)-cuparene, γ-terpinene, dibutyl phthalate, p-cymene, 2-nonen-1-ol, 1,2-epoxy-5-methylhexane, butylated hydroxytoluene and toluene were identified as significant volatile markers in fresh stems (Fig. 5E), in which γ-terpinene and p-cymene with anti-inflammatory and antioxidant effects had synergistic antibacterial effect(Sousa et al., 2022). Only 2-methoxyethanol was screened as volatile marker of the fresh leaves (Fig. 5E). Most of the volatile markers of fresh P. frutescens had antibacterial effects, which may be beneficial for the development of functional food related to oral cavity. Meanwhile, dried leaves had volatile markers of 3-hexen-1-ol, benzaldehyde and trans-2-hexenal, dried seeds had β-copaene, copaene, eudesma-4(14),11-diene and (1S,4S,4aS)-1-isopropyl-4,7-dimethyl-1,2,3,4,4a,5-hexahydronaphthalene, and dried stems had 2-nonen-1-ol, cis-4-thujanol, (1R,4R,5S)-1,8-dimethyl-4-(prop-1-en-2-yl)spiro[4.5]dec-7-ene, p-cymene, 1,2-epoxy-5-methylhexane, tetradecane, 3-octanol and 2-methoxyethano(Fig. 5F). P-cymene exhibited remarkable antiproliferative potential against several cancer cell lines such as human cervical cancer, human non-small cell lung cancer (NSCLC) A549 and A427 cell lines(Balahbib et al., 2021). These potential volatile markers provided the reference for studying differences in therapeutic effect.