Introduction

Perilla crop (Perilla frutescens L.) is an annual self-fertilizing plant of the Lamiaceae family, and it is divided into two cultivated types based on their morphology and uses in East Asia: Perilla frutescens var. frutescens used as an oil or vegetable crop and Perilla frutescens var. crispa used as herbal medicine or a vegetable crop [1,2,3,4,5,6]. In their morphological characteristics, the cultivated type of var. frutescens (CF) has mostly large and soft seeds, green leaves and stems, and non-wrinkly leaves. In contrast, the cultivated type of var. crispa (CC) has small and hard seeds and purple or green coloration on the leaves and stems. These two cultivated types (CF, CC) of Perilla crop are extensively cultivated and used in South Korea, China, and Japan [3,4,5, 7]. In South Korea, CF is used as a leafy vegetable crop and for its seed oil while CC is not cultivated because of its declining use as a herbal medicine. However, it is occasionally found as a relict form in South Korea [3, 4]. Moreover, weedy types of Perilla crop are commonly found in South Korea in places such as wasteland, roadsides, and around farmers’ houses or fields [3, 8]. Although the two cultivated types of Perilla crop (TCTPC) can be well distinguished by morphological characteristics such as leaf color and seed size, it is difficult to distinguish between cultivated and weedy types of Perilla crop based on morphological traits because they have the same number of chromosomes (2n = 40) and there are intermediate types resulting from natural hybridization [9,10,11,12].

Essential oils are naturally occurring volatile compounds with unique odors and are a complex mixture of secondary metabolites that are synthesized and produced by aromatic and spice plants [13, 14]. Volatile compounds, together with sugars and acids, are the main chemical compounds that determine the characteristic aroma and flavor of foods [15]. The essential oil of Perilla leaves contains around 150–200 different compounds, which have high antioxidant, anticancer, anti-inflammatory, insecticidal, and antimicrobial activities [16]. The pharmacological activities of Perilla crop are caused by the content and composition of volatile compounds, which are determined by the chemotype, geographical location, growing environment, growth stage, collection zone, and extraction method [17, 18]. Perilla leaves are an important source of volatile compounds and a rich source of perilla aldehyde (PA), perilla ketone (PK), β-caryophyllene, benzaldehyde, and limonene [19] Depending on the varying content of these compounds, Perilla plants are categorized into seven chemotypes based on the volatile compounds present in their leaves: PA, PK (isoegomaketone), elsholtzia ketone (EK), citral (C), perillene (PL), piperitenone (PT), and phenylpropanoid (PP; myristicin, dill apiole, elemicin) [20,21,22]. Among these compounds, PA is the most abundant compound in CC, and it is used as a food additive for flavor and spiciness and in various medicinal applications for its potent antimicrobial properties against both Candida albicans and airborne microbes [18, 23,24,25]. Moreover, PK is also a major compound for flavor activity in CF, which has potent acute toxicity and cardio-pulmonary toxicity for animals and livestock [26]. Myristicin is a major compound of the CC that is collected in South Korea [27]. It is widely accepted that elemicin and myristicin are psychoactive compounds [16]. In addition, elemicin and dill apiole have been found to be effective in inhibiting proinflammatory cytokines during lung inflammation [16]. Myristicin and dill-apiol are most potent in hexobarbital induced sleeping time prolongation observed in mice [28]. A previous study identified 1-octen-3-ol compound as a key compound in Korean perilla leaves [29]. Another compound, (Z,E)-a-farnesene, has been identified as a major compound in both Chinese CF and CC [18].

Differences in volatile compounds can result in varying physiological effects, and these differences can impact the utilization of the compounds in food and pharmaceutical products [30]. To develop new Perilla varieties with high levels of useful volatile compounds for health, it is essential to conduct research on the content and composition of the volatile compounds in breeding materials. Therefore, we report on the variability of volatile compounds in three different types [CF, weedy type of var. frutescens (WF), weedy type of var. crispa (WC)] of Perilla crop collected from South Korea.

Materials and methods

Plant materials for volatile profile analysis

Information on the number of accessions and their locations for a total of 80 Perilla samples collected in northern, central, and southern regions of South Korea are shown in Table 1 and Addition file 1: Fig. S1. These accessions comprised 40 samples of CF, 20 samples of WF, and 20 samples of WC. This study tentatively categorized the Perilla samples as either the cultivated or weedy type based on a previous study [3]. All of the CF accessions were landraces collected from farms in the various regions in South Korea. In addition, WF and WC were collected on roadsides, in waste areas, and around farmers’ fields or farmhouses in South Korea. In 2022, we cultivated 80 accessions of Perilla crop at the experimental farm of Kangwon National University in Chuncheon, Gangwon-do, and collected leaf samples of each accession during the flowering period.

Table 1 Perilla accessions from different area of South Korea used for volatile compounds profiling
Full size table

Profiling of volatile compounds using gas chromatography-mass spectrometry ( GC–MS)

Volatile compounds were analyzed using a headspace solid-phase microextraction (HS-SPME) injector-equipped Trace 1310 gas chromatograph and TSQ 8000 triple quadrupole mass spectrometer (Thermoscientific, San Jose, CA, USA) and TriPlus RSH autosampler (Waltham, MA, USA) with a DB-Wax capillary column (60 m × 0.25 mm, 0.50 μm; Agilent Technologies). For analysis of the volatile compounds, leaves of CF, WF, and WC were harvested and ground using a Freezer/Mill 6875 (SPEX SamplePrep, Metuchen, NJ, USA). The ground leaves (0.5 g) for each sample were incubated for 10 min at 60 °C. Headspace volatiles of the Perilla leaf samples were adsorbed by inserting SPME fiber assemblies (Supelco, PDMS/DVB; Sigma–Aldrich, St. Louis, MO, USA) into the vial for 10 min at 60 °C with agitation. The injector temperature was 250 °C. The temperature of the gas chromatograph oven was initially maintained at 40 °C for 2 min, ramped to 150 ºC at 4 ºC/min, held for 10 min, then ramped to 200 °C at 4 °C/min, held for 5 min, and finally ramped to 230 °C at 10 °C/min and held for 5 min. Total run time was 65 min, and the helium carrier gas flow was maintained constant at 1.5 mL/min in the splitless with purge mode during the analysis. Total 41 volatile compounds from the leaf samples were identified based on the MS database (NIST version 2.0) with a similarity index (SI) higher than 70%. Similarity index less than 70% was not considered. Quantitative analysis of the percentage of each volatile compound was calculated using the formula below:

$$ {\text{Relative content = M/N * 100\% }} $$

where M is the peak area of the individual volatile compounds and N is the identified total peak area.

Statistical analysis

The relative content of the identified compounds was employed for analysis. Analysis of variance (ANOVA), correlation analysis, principal component analysis (PCA), hierarchical cluster analysis (HCA), and K-means clustering of gas chromatography–mass spectrometry (GC–MS) data were conducted using MetaboAnalyst 5.0, a web-based platform [31]. The significance of differences in the mean values for the various Perilla types was evaluated using Fisher’s LSD test (p < 0.05). To calculate the loading values of eigenvectors and identify the major statistically different components among the samples, PCA was employed. The goal of K-means clustering is to generate k clusters in such a way that the total squared distance between each point and its assigned cluster center is minimized. A dendrogram was generated using Ward’s method of hierarchical clustering and Euclidean distance measurement between the analyzed samples, through HCA analysis of the GC-MS data.

Results

GC–MS Profiling of Volatile Compounds of Perilla leaf

Volatile profiles of 80 Perilla accessions collected in South Korea were analyzed by GC–MS and 41 volatile compounds were identified. These volatile compounds belonged to nine chemical classes: six alcohols, seven aldehydes, two benzodioxoles, two esters, three ethers, four ketones, five monoterpenes, one phenylpropanoid, and eleven sesquiterpenes (Table 2, Additional file 2: Tables S1 and S2, Fig. 1). Ketones comprised 61.7% of the identified compounds, with a predominance of PK in all Perilla accessions. Aldehydes represented approximately 9.7% of the total identified compounds, with the dominance of PA (Table 2; Fig. 1).

Fig. 1
figure 1

Bar plot of ratio of each chemical class for the identified volatile compounds in Perilla accessions. CF: cultivated type of var. frutescens; WF: weedy type of var. frutescens; WC: weedy type of var. crispa

Full size image
Table 2 Composition and comparison of volatile compounds of three types of Perilla accessions
Full size table

This study also investigated the characteristics of the volatile compounds of the three subgroups of Perilla. The highest number of compounds (41) was identified in WC. The aldehydes and benzodioxoles were the most prominent, representing 32.5% and 20.0%, respectively, in WC. A total of 39 compounds were identified in WF with 68.1% being ketones and 12.3% being benzodioxoles. The lowest number of compounds was detected in CF with 34 compounds, which included a high proportion of ketones (87.5%) (Table 2, Additional file 2: Table S2, Fig. 1). Among these compounds, the predominant compound of CF was PK (87.2%). The top three compounds of WF were PK (64.5%), myristicin (7.0%), and dill apiol (5.3%). The main components of WC were PA (26.4%), dill apiol (13.8%), (Z,E)-α-farnesene (9.2%), 1-octen-3-ol (6.2%), and myristicin (5.7%) (Table 2, Additional file 2: Table S1).

Significant differences and correlation analysis of volatile compounds in Perilla crop

This study confirmed significant differences among the three different types (CF, WF, WC) of Perilla accessions for the 41 volatile compounds (Table 2). No significant differences were found among the three types of Perilla accessions for myristicin, methyl salicylate, 3-(4-methyl-3-pentenyl)furan, elemicin, and (E)-β-farnesene. Between WF and CF, five compounds (1-penten-3-ol, 2-penten-1-ol, hexanal, dill apiol, and linalool) exhibited significant differences, while between WF and CF isoegomaketone showed a significant difference. Moreover, between WC and WF and between WC and CF 24 compounds (1-hexanol, (Z)-hex-3-en-1-ol, 1-octen-3-ol, shisool, 2-methylbutanal, β-methylbutanal, E-2-pentenal, (E)-2-hexenal, 2,4-hexadienal, PA, methyl benzoate, 2-ethylfuran, 1-penten-3-one, 6-methyl-5-heptene-2-one, β-pinene, β-myrcene, D-limonene, α-terpineol, eugenol, γ-elemene, bicyclogermacrene, caryophyllene oxide, (Z,E)-α-farnesene, (E)-γ-bisabolene) showed significant differences. One compound, copaene, exhibited significant differences between WC and CF, as well as between WF and CF. Another compound, β-bourbonene, showed significant differences between WF and CF and between WC and CF. The compounds caryophyllene, α-humulene, nerolidol, and PK also showed significant differences among the three types of Perilla accessions (Table 2).

This study evaluated correlation coefficients among the 41 volatile compounds in the 80 Perilla accessions (Fig. 2). Correlation analysis was performed to confirm genetic relationships among the various volatile compounds. Among all 820 combinations, 78 combinations showed comparatively higher positive correlation coefficients (over 0.800 at p < 0.01) than the other combinations. Moreover, two combinations showed relatively higher negative correlation coefficients (less 0.7), between PK and nerolidol (− 0.725 at p < 0.01) and between PK and caryophyllene (− 0.711 at p < 0.01).

Fig. 2
figure 2

Correlation heatmap of the 41 volatile compounds in 80 Perilla accessions

Full size image

PCA and HCA of GC–MS profiling data

To confirm the differences or relationships among the Perilla accessions and to examine which identified volatile compounds were responsible for the major contribution to the difference in Perilla frutescens var. frutescens and var. crispa, PCA and HCA of the detected compounds were performed. The PCA results demonstrated that the first and second principal components accounted for 86.3% of the variance [PC1 (72.5%), PC2 (13.8%)] of all identified volatile compounds in the Perilla accessions (Fig. 3A). PK exhibited highly negative loading scores (− 0.933) on PC1, which enabled the differentiation of all CF and almost all WF accessions from the remaining ones. Volatile compounds with highly positive or negative scores on PC2 were PA (0.731), dill apiol (− 0.623), and myristicin (− 0.245), which enabled all WC and a few WF accessions to be distinguished (Fig. 3B, Additional file 2: Table S3). When K-means clustering was performed with K = 3 on a dataset of the 80 Perilla accessions (Table 3), three clusters were obtained. Cluster 1 consisted of nine WC and four WF accessions, while Cluster 2 comprised all 40 CF, 16 WF, and one WC accession. Cluster 3 contained only 10 WC accessions.

Fig. 3
figure 3

PCA score plot for volatile compounds among 80 Perilla accessions (A), and loading plot for volatile compounds explaining the variation on PC1 and PC2 axes (B)

Full size image
Table 3  K-mean clustering in K = 3 of Perilla accessions using volatile compounds
Full size table

HCA was performed to understand the relationships between the analyzed Perilla accessions. According to the dendrogram and heatmap obtained (Figs. 4 and 5), all Perilla accessions could be divided into two main groups (Group I and II), and these correspond to var. frutescens and var. crispa except for five accessions (PF20-095, PF20-085, PF21-010, PF20-106, PF20-112) (Figs. 4 and 5). In detail, Group I contained 23 accessions with four WF and 19 WC, while Group II included 40 CF, 16 WF, and one WC (Figs. 4 and 5).

Fig. 4
figure 4

Dendrogram obtained by hierarchical cluster analysis on volatile components of 80 Perilla accessions

Full size image

Discussion

Growing demand for functional foods among consumers has driven a greater research emphasis on investigating the physiological and pharmacological properties of compounds derived from plants. Perilla exhibits significant pharmacological activities, including antioxidant, anti-inflammatory, antimicrobial, antidepressant, anxiolytic, chemo-preventive, and antitumor-promoting effects [32,33,34], The presence of bioactive compounds, such as flavonoids, essential oils (volatile compounds), unsaturated fatty acids, triterpenes, and phenolic compounds in Perilla plants is responsible for their pharmacological activities [26]. Perilla’s volatile compounds in particular are considered key components that contribute to its medicinal properties, aroma, and flavor [18, 35]. Therefore, they are commonly utilized in the manufacture of pharmaceuticals and food additives [36, 37].

This study identified a total of 41 volatile compounds belonging to nine chemical classes in this population using GC-MS (Table 2, Additional file 2: Table S1). Although all 41 compounds were identified in 20 accessions of WC, seven (shisool, PA, myristicin, dill apiol, elemicin, β-pinene, α-terpineol) compounds were not detected in CF (Table 2). Comparing the statistical differences of commonly identified compounds between CF and WC, 29 volatile compounds showed significant differences in contents between the two types (Table 2). This suggests that different types of Perilla may exhibit varying aromas, flavors, and pharmacological effects. Both the green-leafed CF and the purple-leafed CC are generally recognized as separate species by botanists [18], and their Korean names are “Deulggae” or “Kkaennip” and “Chajoki” or “Soyup”, respectively. In South Korea, Perilla leaves of CF are mainly used as leafy vegetables, while those of CC are used for medicinal purposes. Moreover, the green-leafed Korean variety of Perilla was the least bitter with a high intensity of cooling sensation, while the red-leaved Perilla was the least astringent and pungent according to a sensory test by electronic nose sensors [38]. The differences in chemical compounds are reflected in the distinct uses and sensory diversity of CF and WC.

Moreover, the predominant compound of CF accessions, with a very high proportion, was PK (80.1%~92.1%). Then, depending on the accession, the next most prominent compounds were (Z,E)-α-farnesene (0.3%~8.8%), methyl salicylate (0.03%~5.0%), and 1-octen-3-ol (1.1%~4.6%) with relatively low proportions (Additional file 2: Table S1). Based on the seven chemotypes for volatile compounds (PA, PK, EK, C, PL, PT, and PP), all of the CF in this study were of the PK type (PK with no isoegomaketone) (Fig. 5, Additional file 2: Table S1). These results may be an indication that the level of diversity for volatile compounds has been reduced by direct or indirect strong selection by people and genetic drift during the domestication from wild or weedy to cultivated type. A previous study confirmed the presence of volatile compounds among five Korean Perilla cultivars (cv. Namcheon, Bora, Saebora, Dongle1, and Dongle2) with purple leaves. The predominant compound was PK, along with some isoegomaketone [39]. The CF accessions used in this study were all landraces collected from farmers. These results suggest that Koreans may prefer the aroma of PK over other aroma types, such as PA, for vegetable uses.

Fig. 5
figure 5

Heatmap with hierarchical clustering with 41 volatile compounds in Perilla accessions

Full size image

The WC accessions in this study had more variation in volatile compounds compared with the CF accessions, with a combination of various compounds being identified in WC (Fig. 1, Additional file 2: Tables S1 and S2). Although most accessions contained PA and PP types, some accessions were clearly divided into different chemotypes. For example, nine accessions (PF20-099, PF18-011, PF20-101, and PF20-152 for PA type; PF20-091, PF21-028, PF20-094, and PF20-082 for PP type of dill apiol; PF19-114 for PP type of myristicin) have a high relative content of PA (62.5%~70.3%), dill apiol (58.5%~77.3), and myristicin (63.9%) with a ratio of over 50%. Although eight WC accessions (PF20-141, PF21-017, PF21-005, PF18-007, PF16-123, and PF21-018 for PA type; PF17-039 and PF21-023 for PP type of elemicin) had a relative content of PA and elemicin of less than 50%, the predominant compounds were still PA (38.0%~49.8) and elemicin (38.1% and 46.2%). Furthermore, the volatile compounds with a relative content ratio of over 10% in at least one accession were 1-octen-3-ol (15.0–20.9% in three accessions), 2,4-hexadienal (13.9% in one), dill apiol (13.3% in one), methyl benzoate (11.5–19.9% in three), isoegomaketone (16.2% in one), PK (11.6% and 38.1% in two), linalool (13.3% in one), (Z,E)-α-farnesene (10.3–21.0% in seven), eugenol (12.0% in one), and caryophyllene (10.4% in one). The accession PF21-012, which is clustered with CF accessions using different clustering methods, contained high ratios of several compounds, including PK (38.1%), isoegomaketone (16.2%), (Z,E)-α-farnesene (15.6%), D-limonene (9.3%), caryophyllene (5.3%), linalool (4.2%), and 1-octen-3-ol (3.7%) (Additional file 2: Table S1). Therefore, this accession can be classified as PK with isoegomaketone type.

Few studies have been conducted to identify the volatile components of WF. This study confirmed the presence of volatile compounds in WF, and among the 41 compounds detected in WC two compounds (elemicin and α-terpineol) were not detected in WF (Table 2). WF is intermediate in morphology compared with CF and WC [3]. It has green leaves and stems like CF and smaller and harder seeds like WC. When comparing the volatile compounds among the three types of Perilla, only two compounds were absent in WF, whereas seven compounds among all the volatile compounds were absent in CF (Table 2). According to ANOVA, 27 volatile compounds showed significant differences in their content between WF and WC, and six volatile compounds exhibited significant differences in their content between WF and CF (Table 2). These results suggest that WF still shows intermediate characteristics in the composition of volatile compounds, and it is difficult to distinguish it from CF and WC based on both morphological characteristics and volatile compounds. This tendency can also be observed in the results of PCA, K-means clustering, HCA, and heatmap analyses, which showed a clear distinction between CF and WC (Table 3; Figs. 3 and 4, and 5). However, WF accessions were not clearly separated from CF or WC, i.e., some WF accessions (PF20-095, PF20-106, PF20-085, and PF21-010) were grouped with other WC accessions (Table 3; Figs. 3 and 4, and 5) while the remaining 16 WF accessions were grouped together with the CF accessions.

In this study, WF accessions exhibited greater variability in volatile compounds compared with CF accessions, while showing less variability compared with WC accessions (Fig. 1, Additional file 2: Tables S1 and S2). Seventeen accessions (PF20-005, PF20-011, and PF20-026 for PK type with isoegomaketone; PF20-137, PF20-024, PF20-039, PF20-083, PF20-142, PF20-098, PF20-153, PF20-115, PF20-121, PF20-103, PF20-149, PF19-133, and PF20-112 for PK type with no isoegomaketone; PF20-095 for PP of myristicin) had a high relative content of PK (52.3%~89.8%) and myristicin (77.5%) with a ratio of over 50%. Although three accessions (PF20-106 and PF21-010 for PP type of dill apiol; PF20-085 for PP type of myristicin) had a relative content of PA and elemicin of less than 50%, the predominant compounds were dill apiol (36.7% and 46.8) and myristicin (46.9%). Furthermore, the volatile compounds with a relative content ratio of over 10% in at least one accession were 1-octen-3-ol (11.1 in one accession), dill apiol (22.7% in one), methyl salicylate (13.6% in one), isoegomaketone (12.1%~31.2% in three), (E)-β-farnesene (11.0% in one), and (Z,E)-α-farnesene (11.6% in one). WC and WF are not cultivated by farmers in South Korea, and they are often found growing around farmers’ houses and fields of CF [3, 4]. Natural hybridization between weedy and cultivated types and introgression of many useful genes makes them adapted to natural growing conditions [40]. These complex content and composition patterns in WF may be the result of natural hybridization between WF and CF or WF and WC, which might have originated from multiple sources. Currently, there are no precise reports on the rate of natural hybridization among cultivated types of Perilla and its weedy types. However, previous studies have recently reported successful artificial hybridization between var. frutescens and var. crispa [12, 41, 42].

The composition of compounds can vary depending on the plant part (leaves, stems, seeds), treatment (fresh or dried), growth stage (seedling or mature), geographical location, etc. [17, 18, 43, 44]. The volatile compounds present in the fresh leaves at the heading period of CF, WF, and WC were collected from various regions (northern Gyeonggi-do, central Chungcheongnam-do, southern Gyeongsangnam-do, Jeollabuk-do, and Jeollanam-do) of South Korea. This study conducted a correlation test between the 41 volatile compounds and the collecting regions, and this revealed statistically significant differences in a few compounds (isoegomaketone with r2 = − 0.350 at p < 0.01, 1-octen-3-ol with r2 = 0.261, and methyl benzoate with r2 = 0.220 of at p < 0.05) (Data not shown). Although these compounds have relatively low correlation coefficient values, three out of four accessions (PF21-025, PF21-027, and PF21-028) collected from the northern part of South Korea contained isoegomaketone (12.1%~31.2%) (Table 1, Additional file 2: Table S1). However, through further research analyzing the accessions from excluded regions (such as Gangwon-do and other cities in the same area as those in this study), it is expected that more distinct regional patterns can be revealed. A previous study of CF and CC in China detected volatile compounds at the microgreen (seedling) stage [17]. Although this previous study was performed using only two Perilla cultivars, it showed that Chinese CC with purple leaves has D-limonene as the main compound with low levels of PA, while Chinese CF with green leaves has 2-hexanoylfuran as the main compound with a high ratio of perillene. Another previous study [18] confirmed PA and limonene as the main compounds in CC (Zisu), while PK was identified as the main compound in CF (Baisu) in China. These findings are similar to the results of our study on volatile compounds. The results indicate that differences and similarities could be attributed to variations in collecting region, growth stage, number of accessions, customer preferences, and cultivation environment [17, 18, 43, 44].