Introduction

Perilla frutescens Britt. (Labiatae) is a perennial crop that performs self-pollination and grows intensively in Himalayan mountain areas in Southeast Asia and East Asia. Recently, it has been introduced to Europe and North America due to its economic properties as an oil crop or a garden plant (Nitta et al. 2003). In East Asia, Perilla crop is cultivated in a large scale and used widely; therefore, East Asia is considered to be the birth place of P. frutescens (Makino 1961; Li 1969; Nitta 2001). China has been assumed to be the primary center of biodiversity for Perilla crop (Li 1969; Zeven and de-Wet 1982). Lee and Ohnishi (2001, 2003) suggested that Korea is the secondary center of diversity of cultivated var. frutescens due to its large-scale cultivation, various usages, and higher level of morphological diversity.

Perilla crop has been divided into two cultivated types, P. frutescens var. frutescens and P. frutescens var. crispa, based on their morphology and use. P. frutescens var. frutescens is a kind of oil crop and called by different names in East Asian countries, such as Ren in China, Dlggae in Korea, and Egoma in Japan. It is also used as a leafy vegetable crop only in Korea, and its seeds were traditionally used in the same way with sesame seeds in China, Korea, and Japan from old times (Lee and Ohnishi 2001, 2003; Nitta et al. 2003). Meanwhile, P. frutescens var. crispa is a Chinese medicine crop that is called Cha-jo-ki in Korean, Shiso in Japan, and Zisu in China (Lee and Ohnishi 2001, 2003). It is also used as a spicy vegetable or pickle crop in Japan. Thus, these two cultivated types of Perilla crop have been important in East Asia from ancient times (Lee and Ohnishi 2003; Nitta et al. 2003).

In China, Perilla crop can be found in multiple provinces, such as Heilongjiang, Liaoning, Shanxi, Ningxia, Gansu, Anhui, Hubei, Sichuan, Taiwan, Hunan, Jiangxi, Jiangsu, Anhui, Henan, and Hulunbuir city of Inner Mongolia (Liu et al. 1996; Liu and Zhang 1998; Zhang et al. 2009; Hu et al. 2010). The main producing areas for P. frutescens var. crispa are concentrated in Guizhou and Sichuan provinces (Zhang et al. 2009). The weedy forms of the two cultivated types of Perilla crop are grown and commonly found in such habitats as roadsides, the edge of village and around farmers’ fields or farmhouses. In northern areas of China, the cultivated area of P. frutescens var. frutescens is large in Changchun, Jilin, and Songyuan of Jilin province. Particularly, Yanbian area of Northeast China, inhabited by Chinese and Korean, var. frutescens is cultivated in a larger scale. There, the seeds of var. frutescens are used for flavoring in traditional foods, Perilla seed oil had been used for vegetable oil or industrial purpose, while cultivated and weedy types of var. crispa were found all over the provinces in South China. In South China, var. crispa is used majorly due to its medicinal function (Tan et al. 2012; Wang and Guo 2012; Wei et al. 2015). For instance, leaves of var. crispa exhibit the function of detoxification and have been used in cooking crab and fish for more than 2000 years (Yu et al. 2016). In addition, seeds and leaves of var. crispa have been considered effective in the treatment of cough, common cold, asthma, and digestive problem (Yu et al. 2016).

Taxonomic studies have underpinned the management of genetic resources in many aspects. Traditionally, the researches on plant taxonomy are conducted based on the comparative analysis in morphological traits. For example, morphological traits of Perilla accessions collected from East Asian areas were investigated by Lee and Ohnishi (2001) and Ma and Lee (2017) for the better understanding of morphological and geographical variation among different types of Perilla crop. They found many traits such as leaf size, seed size, plant height, branch number, flower, leaf and stem color, degree of pubescence, and plant fragrance that can be used for the discrimination of var. frutescence and var. crispa. Among these morphological traits, seed size can be used as the most reliable trait for separating cultivated type of var. frutescence from var. crispa and weedy var. frutescence. However, morphological analysis can be impacted by the environmental factors and provide the dubious results of taxonomy (Rao and Hodgkin 2002). Significant differences were not found between the cultivated and weedy var. crispa in morphological characters (Lee and Ohnishi 2001). Some accessions of var. frutescens performed similar qualitative traits (color of leaf, flower and stem) with var. crispa (Lee and Ohnishi 2001; Ma and Lee 2017), which created difficulties in the classification among Perilla accessions. The application of molecular technologies in taxonomic analysis has remedied the shortages of analysis based on morphological traits and has provided new insights into the phylogeny and taxonomy of many plant species. A number of excellent properties for molecular techniques have been demonstrated by Palmer et al. (1988). They reported that molecular techniques can reduce the impact of environment on the final data and analyze more characters than morphological analysis.

Therefore, various molecular markers have provided useful information regarding genetic diversity and genetic relationships in many crops (Senior et al. 1998; Nitta and Ohnishi 1999; Prasad et al. 2000; Lee et al. 2002; Hamza et al. 2004; Xia et al. 2005; Lee and Kim 2007; Ganapathy et al. 2011; Park et al. 2008, 2015, Sa et al. 2013, 2015; Zhang et al. 2016; Ma et al. 2017). Among them, random amplified polymorphic DNA or RAPD (Nitta and Ohnishi 1999), amplified fragment length polymorphism or AFLP (Lee et al. 2002; Lee and Ohnishi 2003), and simple sequence repeats or SSR (Lee and Kim 2007; Park et al. 2008, Sa et al. 2013, 2015; Ma et al. 2017) markers have been applied in the analysis of genetic diversity and relationships among cultivated and weedy types of Perilla crop in East Asia and other countries. Particularly, SSR markers are highly reproducible, polymorphic, generally codominant, and abundant in plant genomes (Powell et al. 1996; Park et al. 2009). As a result of these better features, SSRs have been used to establish genetic diversity and genetic relationships in many crop species (Prasad et al. 2000; Da Cunha et al. 2014; Yook et al. 2014; Park et al. 2015).

Identification of genetic variation is an essential ability for the long-term success of breeding programs and maximizes the use of germplasm resources (Xia et al. 2005; Park et al. 2008). In East Asia, China has a long history of the cultivation of Perilla crop, but there has been little research on the genetic diversity and genetic relationships among accessions of Perilla crop and their weedy types. Therefore, in this study, we used 21 Perilla SSR primers in order to determine the genetic diversity, genetic relationships, and population structure of some collected Perilla accessions in the northern and southern areas of China.

Materials and methods

Plant materials

A total of 91 accessions (54 cultivated var. frutescens, 27 weedy var. frutescens, 4 cultivated var. crispa, and 6 weedy var. crispa) were collected from northern and southern areas of China and deposited in the National Agrobiodiversity Center, Rural Development and Administration, Jeonju, Republic of Korea, for permanent seed preservation. The number of accessions and the name of the collection place for each Perilla accessions have been shown in Table 1 and Fig. 1.

Table 1 Perilla accessions from different regions of China used for SSR analysis
Full size table
Fig. 1
figure 1

Collecting sites of 91 Perilla accessions collected from Northern and Southern China. The two circles receptively showed the Perilla accessions collected in the northern (upper) and southern (lower) regions of China. The small map shown by the arrow is an enlargement of the Yunnan Province area in China

Full size image

DNA extraction, SSR analysis, and silver staining

Genomic DNA was extracted from the young leaves for each Perilla accession at the seedling stage following the Plant DNAzol Reagent protocol (GibcoBRL Inc., Grand Island, NY, USA). The Perilla SSR primers used in our experiments were developed and used by Kwon et al. (2005), Park et al. (2008), Song et al. (2012), and Sa et al. (2013, 2015). SSR amplifications were conducted in a total volume of 20 µL, consisting of 20 ng genomic DNA, 1X PCR buffer, 0.5 µM forward and reverse primers, 0.2 mM dNTPs, and 1 U Taq polymerase (Biotools, Spain). The PCR profile consisted of initial denaturation at 95 °C for 3 min, followed by 36 cycles of 95 °C for 30 s, 55 °C for 30 s, and 72 °C for 1 min 30 s, with a final extension step of 5 min at 72 °C. After PCR, 5 µL of the final products were mixed with 10 µL electrophoresis loading buffer (98% formamide, 0.02% BPH, 0.02% Xylene C, and 5 mM NaOH). After denaturing and quick cooling, 2 µL of each sample was loaded onto 6% denaturing (7.5 M urea) acrylamide-bisacrylamide gel (19:1) in 1X TBE buffer and then electrophoresed at 1800 V and 60 W for 130 min. The separated fragments were then visualized through a silver staining kit (Promega, Madison, WI, USA).

Data analysis

Fragments amplified using the SSR primers were scored as present (1) or absent (0). Power Marker version 3.25 (Liu and Muse 2005) was applied to obtain the information on the number of alleles, allele frequency, major allele frequency (MAF), gene diversity (GD), and polymorphic information content (PIC). Genetic similarities were calculated for each pair of lines using the Dice similarity index (Dice 1945). To illustrate the genetic relationship of samples, the similarity matrix was then used to construct an unweighted pair group method with arithmetic mean (UPGMA) dendrogram by the application of SAHN-Clustering from NTSYS-pc.V.2.1 (Rohlf 1998). The STRUCTURE 2.2 software (Pritchard and Wen 2003) was used for the investigation of a population structure for 91 accessions of Perilla crop. To identify the number of clusters (K), the software was run using admixture model with correlated allele frequencies following Hardy–Weinberg equilibrium. Five independent runs were performed for each simulated value of K = 1–10, with a burn-in period of 100,000 runs followed by 100,000 Monte Carlo Markov Chain (MCMC) replications. The real number of population was detected by plotting Ln (PD) derived ΔK, an ad hoc quantity against each K (Evanno et al. 2005). The result was computed and visualized using the Structure Harvester software (http://taylor0.biology.ucla.edu/struct_harvest/).

Results

SSR polymorphism and genetic variation between Perilla accessions from northern and southern areas of China

The 21 SSR primer pairs with polymorphism were applied for final detection in this study. The information on primer pairs, allele size range, number of alleles, MAF, and values of GD and PIC was shown in Table 2. A total of 147 alleles were produced among 91 Perilla accessions. The average number of alleles revealed by each primer pair was 7, varying from 3 for GBPF201 and KWPE56 to 12 for KWPE57. The MAF for each SSR marker ranged from 0.228 at GBPF135 to 0.957 t GBPF201, with the average value of 0.549. The values of GD ranged from 0.084 revealed by GBPF201 to 0.828 revealed by KWPE57 with the average value of 0.577. Several higher values of genetic diversity were also observed at KWPE53 (0.783), KWPE19 (0.776), GBPF203 (0.741), GBPF135 (0.827), GBPF204 (0.804), and GBPF111 (0.762), respectively. Allelic diversity and frequency among these Perilla accessions could be reflected by the PIC values, which ranged from 0.082 at GBPF201 to 0.808 at KWPE57, with an average value of 0.537. 14 markers—GBPF179, GBPF135, KWPE25, KWPE26, GBPF203, KWPE39, KWPE57, KWPE53, KWPE58, KWPE19, KWPE29, GBPF111, GBPF155, and GBPF204—exhibited PIC values higher than 0.5.

Table 2 Characteristics of the 21 SSR loci including primer sequence, repeat motif, allele size range, no. of alleles, gene diversity, and polymorphic information content among 91 accessions from northern and southern areas of China
Full size table

To compare the genetic diversity of Perilla accessions from Northern China (45 accessions) and Southern China (46 accessions), the number of alleles, values of genetic diversity, and polymorphic information content of Perilla accessions from Northern and Southern China were detected respectively and summarized in Table 3. The average numbers of alleles per locus were 5.29 and 5.81 for Perilla accessions from northern and southern areas of China, respectively. The average values of GD were 0.462 and 0.566 for Perilla accessions from northern and southern areas of China, respectively. The average values for PIC were 0.432 and 0.526 for Perilla accessions from northern and southern areas of China, respectively. Obviously, the genetic diversity level of Perilla accessions from southern areas of China was higher than those accessions from northern areas of China. As shown in Table 4, average numbers of alleles per loci were 4.81 and 4.10 for accessions of cultivated var. frutescens from northern and southern areas of China, respectively. The average values of GD were 0.400 and 0.532 for accessions of cultivated var. frutescens from northern and southern areas of China, respectively. The average values for PIC were 0.374 and 0.490 for accessions of cultivated var. frutescens from northern and southern areas of China, respectively. This indicated that the genetic diversity for accessions of cultivated var. frutescens in Southern China was higher than that in Northern China. In Table 5, the average values for the number of alleles, GD, and PIC for accessions of cultivated var. frutescens were 6, 0.467, and 0.439, respectively. The average values for the number of alleles, GD, and PIC for accessions of weedy var. frutescens were 4.71, 0.500, and 0.460, respectively. The average values for the number of alleles, GD, and PIC for accessions of var. crispa were 4.05, 0.567, and 0.524, respectively. That implied that accessions of var. crispa performed the highest level of genetic diversity among different types of Perilla accessions in China.

Table 3 No. of alleles, genetic diversity, and polymorphic information content (PIC) obtained from each SSR locus in Perilla accessions from northern and southern areas of China
Full size table
Table 4 No. of alleles, genetic diversity, and polymorphic information content (PIC) obtained from each SSR locus in accessions of cultivated var. frutescens from northern and southern areas of China
Full size table
Table 5 No. of alleles, genetic diversity, and polymorphic information content (PIC) obtained from each SSR locus in accessions of Perilla crop and their weedy types from China
Full size table

Genetic relationships among accessions of Perilla crop from different regions of China

The phylogenetic tree was constructed using UPGMA, and 91 Perilla accessions were revealed to have clustered into 3 major groups (Group I, Group II, and Group III) with a genetic similarity of 38% (Fig. 2). Group I could be further divided into four subgroups under a genetic similarity of 46%. Group I-1 contained 36 accessions of cultivated var. frutescens from Northern China and 2 accessions (CSY21 and CSY22) of cultivated var. frutescens from Southern China. Group I-2 included seven cultivated accessions of var. frutescens and one accession of weedy var. frutescens from Southern China. Group I-3 consisted of two accessions of cultivated var. frutescens from Northern China. Group I-4 was comprised of three accessions of cultivated var. frutescens from Northern China and one accession of cultivated var. crispa (CH53) from Northern China. Group II was further classified into four subgroups under a genetic similarity of 43%. Group II-1 contained 21 accessions of weedy var. frutescens, 5 accessions of weedy var. crispa, and 1 accession (CSY4) of cultivated var. frutescens from Southern China. Group II-2 included 3 accessions of cultivated var. frutescens and 2 accessions of weedy var. frutescens from Southern China. Group II-3 consisted of 2 accessions (CSY2 and CSY3) of weedy var. frutescens. Group II-4 was composed of one accession (CSY46) of weedy var. frutescens from Southern China. Group III was comprised of three accessions of cultivated var. crispa from Northern China and one weedy accession (CSY48) of var. crispa from Southern China.

Fig. 2
figure 2

UPGMA dendrogram based on the SSR markers. The accessions of Perilla crop from northern and southern areas of China are shown in Table 1. White circle: cultivated var. frutescens from Northern China, black circle: cultivated var. crispa from Northern China, white square: cultivated var. frutescens from Southern China, black square: weedy var. frutescens from Southern China, black up-pointing triangle: weedy var. crispa from Southern China

Full size image

In our analysis, most Perilla accessions collected in Northern China were clearly distinguished from accessions collected in Southern China. For accessions from Northern China, all accessions of cultivated var. frutescens were clearly discriminated from the cultivated var. crispa, except for one accession (CH53) which was located in the group of cultivated var. frutescens in the phylogenetic tree (Fig. 2). Furthermore, in the case of accessions from Southern China, most accessions of cultivated var. frutescens was differentiated from weedy accessions of var. frutescens, but most accessions of weedy var. crispa were mixed with weedy var. frutescens.

Population structure

To identify the genetic structure among the 91 Perilla accessions from northern and southern areas of China, a model-based approach in the STRUCTURE software was employed to subdivide each accession into its corresponding subgroups. The ad hoc measure ΔK following the method demonstrated by Evanno et al. (2005) was applied to overcome the difficulty in interpreting the real K values. The highest value of ∆K for the 91 Perilla accessions was for K = 2 (Fig. 3).

Fig. 3
figure 3

Magnitude of ∆K as a function of K. The peak value of ∆K was at K = 2, suggesting two genetic clusters in Perilla accessions from northern and southern areas of China. X-axis is K value and Y-axis is ∆K value

Full size image

Clustering bar plots with K = 2 are shown in Fig. 4. At K = 2, all 91 Perilla accessions were divided into Group I, Group II, and an admixed group. There was no clear geographic structure among the 91 Perilla accessions from southern and northern areas of China, which has been revealed in the analysis of UPGMA (Figs. 2, 4). Therefore, we conducted an analysis following the method of Wang et al. (2008) according to a membership probability threshold of 0.8. The result showed that the 91 Perilla accessions were classified into Group I, Group II, and an admixed group. Group I included 32 accessions (CH3, CH35, CH38, CH46, CH62, CH64, CH65, CH66, CH72, CH75, CH79, CH45, CH82, CH83, CH16, CH23, CH32, CH33, CH14, CH42, CH5, CH8, CH34, CH37, CH30, CH39, CH50, CH31, CH6, CH48, CH26, and CH47) of cultivated var. frutescens from Northern China and 5 accessions (CSY18, CSY21, CSY22, CSY31, and CSY40) of cultivated var. frutescens from Southern China. Group II included 3 accessions (CH52, CH2, and CH54) of cultivated var. crispa from Northern China, 1 accession (CSY39) of cultivated var. frutescens, 26 accessions (CSY6, CSY9, CSY10, CSY11, CSY12, CSY13, CSY14, CSY15, CSY19, CSY24, CSY25, CSY28, CSY32, CSY41, CSY42, CSY44, CSY45, CSY46, CSY3, CSY26, CSY27, CSY29, CSY30, CSY2, CSY5, and CSY20) of weedy var. frutescens, and 6 accessions (CSY7, CSY8, CSY16, CSY17, CSY23, and CSY48) of weedy var. crispa from Southern China. The admixed group is comprised of nine accessions (CH24, CH29, CH51, CH44, CH12, CH10, CH9, CH25, and CH36) of cultivated var. frutescens and one accession (CH53) of cultivated var. crispa from Northern China, seven accessions (CSY4, CSY33, CSY34, CSY35, CSY36, CSY37, and CSY38) of cultivated var. frutescens and one accession (CSY1) of weedy var. frutescens from Southern China.

Fig. 4
figure 4

Population structure of 91 accessions from northern and southern areas of China based on 21 SSRs for K = 2. The two lines are the criteria for distinguishing three groups, Group I, Group II, and an admixed group. White circle: cultivated var. frutescens from Northern China, black circle: cultivated var. crispa from Northern China, white square: cultivated var. frutescens from Southern China, black square: weedy var. frutescens from Southern China, black up-pointing triangle: weedy var. crispa from Southern China

Full size image

Discussion

Investigation of the genetic diversity among landraces or cultivars and their geographical distribution of a given crop are one of the basic approaches to the identification of the origin and of varietal differentiation of the crop. In East Asia, China is considered to be one center of diversity of Perilla crop, however, by this time, there is very little study for genetic diversity and genetic relationships on Perilla accessions from northern and southern areas of China. Genetic diversity was attributed to many genetic differences between individuals and may be manifested in differences in DNA polymorphism (e.g., RFLP, RAPD, AFLP, and SSR markers) (Powell et al. 1996; Park et al. 2009). Particularly, utilization of SSR markers for detecting genetic diversity and genetic relationships among landraces or cultivars has been well-established for many crops (Prasad et al. 2000 for Wheat; Ni et al. 2002 for rice; Hamza et al. 2004 for barley; Xia et al. 2005 for maize; Ali et al. 2008 for sweet sorghum; Sa et al. 2013, 2015 for Perilla crop).

In this study, 21 SSR markers were applied for the assessment of genetic diversity among 91 Perilla accessions from southern and northern areas of China. According to our results, a total of 147 alleles were amplified by 21 SSRs from the 91 Perilla accessions from different areas of China, which produced an average of 7 alleles per locus. This value is a little lower than the effective number of alleles per SSR locus obtained in the results of previous studies. Lee and Kim (2007) detected 101 alleles among 70 Perilla accessions from East Asia using 11 SSRs, with an average allele number of 9.2 per locus. Sa et al. (2013) obtained 165 alleles using 18 SSR markers in 56 Perilla accessions from Korea and Japan with an average of 9.2 alleles yielded at each locus. Sa et al. (2015) produced 166 alleles by 18 SSR markers in 81 Perilla accessions from East Asia and other countries with an average of 9.8 alleles in each locus. Recently, Ma et al. (2017) obtained 95 alleles using 21 SSR markers in 77 Perilla accessions from northern areas of China with an average of 4.52 alleles yielded at each locus. In our study, the results showed a lower average value in genetic diversity than in previous studies by Lee and Kim (2007) and Sa et al. (2013, 2015), but showed a higher average value in genetic diversity than in previous studies by Ma et al. (2017). The reason for the low and high values of genetic diversity and allele numbers per locus obtained in the current study might be attributed to the lower or higher proportion of weedy accessions of var. frutescens and the narrow or wide collecting sites compared with the previous studies. For example, most Perilla accessions analyzed in this study are accessions of the cultivated var. frutescens collected from southern and northern areas of China. Unfortunately, weedy Perilla accessions were just collected from one province (Yunnan) in southern China. And also weedy Perilla accessions were not collected from northern areas compare to southern areas of China. In other studies, a higher proportion of weedy types of Perilla crop were used from several provinces of South China, Korea, Japan, and other countries (Lee and Kim 2007; Sa et al. 2013, 2015). These researchers have reported that weedy Perilla accessions performed a higher genetic diversity level than cultivated Perilla accessions and suggested that domestication via direct or indirect selections from human or nature during the evolutionary stage from wild type to cultivated type caused some alleles to be lost, and led to the declined level of polymorphism and genetic diversity (Lee and Ohnishi 2003; Lee and Kim 2007; Sa et al. 2013, 2015). This conclusion was also proven in our study. Perilla accessions from Southern China comprising a high proportion of weedy Perilla accessions showed higher values of average number of alleles, GD, and PIC than Perilla accessions of Northern China consisting of 41 accessions of cultivated var. frutescens and four accessions of cultivated var. crispa (Table 3). A genetic diversity level for 27 accessions of weedy var. frutescens was higher than that of 54 accessions of cultivated var. frutescens (Table 5). Ten accessions of var. crispa exhibited the highest genetic diversity level compared with other types of Perilla accessions (Table 5). In the previous studies of Lee and Ohnishi (2001) and Sa et al. (2012, 2015), it was reported that the cultivated var. frutescens might be differentiable from the weedy var. frutescens but that the cultivated and weedy types of var. crispa would not prove sufficiently differentiable. Because the weedy var. frutescens and cultivated and weedy types of var. crispa had small seeds (less than 2 mm) with seed dormancy, whereas the cultivated var. frutescens had large seeds (larger than 2 mm) which are free from seed dormancy (Lee and Ohnishi 2001; Sa et al. 2012).

Considering seed size between cultivated and weedy types of var. frutescens and var. crispa indicated that further, Lee and Ohnishi (2001, 2003) suggested that the cultivated var. frutescens has been sufficiently differentiated from the weedy var. frutescens, whereas, the cultivated var. crispa has not yet sufficiently differentiated a domesticated form from the weedy or wild var. crispa, as well as previous reported by Lee and Ohnishi (2001, 2003). Although the wild species of the Perilla crop has not yet been reported in East Asia, our results suggest that the weedy types of P. frutescens including cultivated types of var. frutescens and var. crispa are the key taxa to understanding the origin of the cultivated type of var. frutescens, as mentioned in a previous report by Lee and Ohnishi (2001, 2003) and Sa et al. (2013, 2015).

On the other hand, accessions of the cultivated var. frutescens from southern areas of China showed much higher genetic diversity and PIC values than those of the cultivated var. frutescens from northern areas of China (Table 4), although accessions of the cultivated var. frutescens from southern areas of China had a smaller population size and narrow geographical distribution. In a previous report by Nitta et al. (2005), in China, Perilla accessions including cultivated and weedy types were widely distributed in village, roadside, rivers, or mountains. When we collected the Perilla genetic resources for the Yunnan Provinces in Southern China, these accessions of cultivated and weedy types of Perilla crop were founded in rivers or mountain areas. Various plant species collected from mountainous or plateau areas have been reported to perform a higher genetic diversity and variation, such as kiwi (Lai et al. 2015), strawberry (Zhang et al. 2016), cherry (Meng et al. 2015), and sugarcane (Zhang et al. 2013). The higher genetic diversity in cultivated accessions of var. frutescens from Southern China may be attributed to the complex topography and relatively weak interference of human activities in their collecting sites which is located on the Yunnan-Guizhou Plateau. Higher genetic diversity was formed for the adaptation to the higher altitude (1000–3000 m), complex topography, and climate. In contrast, the topography of the collecting sites in northern areas of China was major in the plain, where the environment is relatively softer and tends to be prone to habitat destruction due to industrialization. In addition, gene drift usually occurs under the stress of farmers’ selection and also genetic bottleneck due to the expansion of cultivation from southern to northern regions, which may decrease the genetic diversity of local Perilla accessions in China. Therefore, the genetic diversity level of cultivated accessions of var. frutescens from Northern China was lower in spite of a larger population size and a wider geographical distribution of collecting sites.

The phylogenetic tree was constructed by UPGMA analysis and shown in Fig. 2, 91 Perilla accessions were classified into three main groups. Most cultivated Perilla accessions were clearly separated from weedy accessions, implying that artificial selection led to the genetic differences between cultivated and weedy Perilla accessions. For Perilla accessions in Group I, most cultivated accessions of var. frutescens from Southern China were situated in Group I-2 and obviously separated from cultivated accessions of var. frutescens from Northern China under the genetic similarity of 46%, indicating that gene flow occurring between cultivated accessions of var. frutescens from Southern and Northern China was limited significantly as a result of the long distance between collecting sites in southern and northern areas of China. Meanwhile, ambiguous classification was also found in this phylogenetic tree. For example, one accession (CSY1) of weedy var. frutescens and one accession (CH53) of cultivated var. crispa were grouped with the accessions of cultivated var. frutescens. Most accessions of weedy var. crispa were mixed with the accessions of weedy var. frutescens. Several accessions of cultivated var. frutescens (CSY4, CSY36, CSY37, and CSY39) from Southern China were observed in the group of weedy var. frutescens. Based on the morphological measurement, some special accessions such as CSY2, CSY26, CSY27, CSY29, and CSY30, which emitted a specific fragrance to var. frutescens and exhibited similar morphological traits with var. crispa, such as purple color of flower, leaves, and stem, were found. These accessions were grouped with weedy accessions of var. frutescens in the phylogenetic tree of this study. Therefore, these special accessions should be classified as weedy accessions of var. frutescens. Admixed groups were also reported in the studies of Lee and Ohnishi (2003) and Sa et al. (2013) on the basis of AFLP and SSR markers, respectively. Based on the conclusion of Lee et al. (2002); Lee and Kim (2007), Lee and Ohnishi (2001, 2003), Nitta and Ohnishi (1999), and Sa et al. (2013, 2015), it was assumed that CSY1 might be an escaped form from cultivation with small and hard seeds. CH53 might be the hybrid originating from natural hybridization between var. frutescens and var. crispa. These weedy accessions of var. crispa were classified ambiguously in the phylogenetic tree, which might have originated from the hybrids between these two weedy types of Perilla frutescens. The two accessions (CSY21 and CSY22) of cultivated var. frutescens from Southern China were situated in the group of cultivated var. frutescens from Northern China, which might be attributed to human business activities transferring the seeds of CSY21 and CSY22 from Northern China to Southern China. A population structure analysis in Fig. 4 showed that Group I was only composed of accessions of the var. frutescens from Northern and Southern China. Group II comprised most accessions of weedy var. frutescens and var. crispa from Southern China except three accessions (CH2, CH52, and CH54) of cultivated var. crispa from Northern China and one accession (CSY39) of cultivated var. frutescens from Southern China. However, the geographical locations of these Perilla accessions were not coincidental with their positions in the UPGMA dendrogram and population structure, indicating that the diffusion of Chinese Perilla accessions in China might be conducted through multiple routes and might neither be directional nor gradual, as in a previous report by Lee and Ohnishi (2003).

Development of SSR markers are essential for the investigations on genetic diversity, population structure, and collection and conservation of germplasm resources. In this study, 21 SSR markers have performed high efficiency in the differentiation between cultivated and weedy Perilla accessions, even in the classification of accessions from different areas in China. Valuable information on the genetic structures and relationships of Perilla accessions from northern and southern areas of China will assist breeders to make appropriate strategies for Perilla crop breeding and germplasm conservation.