Genetic variation and association mapping in the F2 population of the Perilla crop (Perilla frutescens L.) using new developed Perilla SSR markers

The transcriptome sequencing approach RNA-seq represents a powerful tool for transcriptional analysis and development of simple sequence repeat (SSR) markers for nonmodel crop. In the Perilla crop, analysis of the distribution of different repeat motifs showed that the most abundant type was dinucleotide repeats (62.0%), followed by trinucleotide repeats (35.3%), with the two together comprising 97.3% of the eSSR repeats. In this study, we developed 39 new SSR primer sets by the transcriptome sequencing approach RNA-sEq. In total, 130 alleles were detected segregating in nine Perilla accessions with an average of 3.3 alleles per locus, ranging from 125 to 360 bp. The number of alleles per locus ranged from two to six. To detect SSR markers associated with morphological characteristics of Perilla crop, 40 individuals from an F2 population of Perilla were selected for association analysis based on their leaf- and plant-related characteristics. An association analysis of 37 SSR markers and 9 leaf- and plant-related traits in the 40 individuals of the F2 population was conducted. From the analysis, we identified 12 SSR markers associated with leaf-related traits and 11 SSR markers associated with plant-related traits. Therefore, the new Perilla SSR primers described in this study could be helpful in identifying genetic diversity and genetic mapping, designating important genes/QTLs for Perilla crop breeding programs, and allowing Perilla breeders to improve leaf and plant quality through marker-assisted selection (MAS) breeding programs.

. These two cultivated types of Perilla crop have a long history of cultivation and are very important as oil crop and vegetable or medicine crop in East Asia Ohnishi 2001, 2003;Nitta et al. 2003). Although var. frutescens and var. crispa can be crossed with each other by artificial pollination, these two cultivated types of Perilla crop have different morphological characteristics (Honda et al. 1990(Honda et al. , 1994Lim et al. 2019). For example, Perilla frutescens var. frutescens is taller with larger seed size (above 2 mm), either soft or hard seeds, green leaves and stems, and non-wrinkled leaves and a fragrance specific to the var. frutescens. Meanwhile, Perilla frutescens var. crispa is shorter with a smaller seed size (below 2 mm), only hard seeds, red or green coloration in the leaves and stem, and wrinkly or nonwrinkly leaves and a fragrance specific to the var. crispa Ohnishi 2001, 2003;Lee et al. 2002;Nitta et al. 2003).
Many taxonomic studies have been conducted to distinguish between the two cultivated types of Perilla crop using morphological characteristics and molecular markers. Their chromosome number is the same, 2n = 40 (Honda et al. 1994;Yamane 1950). A cross experiment between the two cultivated types by Nagai (1935), Honda et al. (1990Honda et al. ( , 1994 and Lim et al. (2019) showed that they are cross-fertile with each other by artificial pollination. Furthermore, because of the existence of intermediate types, studies by Koezuka et al. (1985Koezuka et al. ( , 1986, Honda et al. (1990), and Ohnishi (2001, 2003) failed to distinguish between the two cultivated types of Perilla crop. Essential oils, anthocyanins, and the color and hardness of the seeds cannot be used as key characteristics so that it is still difficult to distinguish the two cultivated types of Perilla crop. In addition, weedy plants of Perilla crop were first reported in the two cultivated types of P. frutescens by Ohnishi (2001, 2003), Nitta and Ohnishi (1999) and Nitta et al. (2003). They reported that analyses of random amplification of polymorphic DNA (RAPD) and amplified fragment length polymorphism (AFLP) markers showed that weedy plants could be grouped into two types: one belonging to the var. frutescens group and another belonging to the var. crispa group. However, the analysis does not provide concrete evidence of the origin of the weedy plants of the Perilla crop. Therefore, further research is needed to distinguish between cultivated and weedy types of Perilla crop.
Understanding the molecular genetic mechanism of phenotypic variation remains a great challenge because most complicated traits, such as yield, quality and resistance, are controlled by quantitative trait loci (QTLs) (i.e. quantitative and qualitative traits) (Mazzcucato et al. 2008). In crop breeding studies, genetic mapping of functional loci for quantitative and qualitative related traits is used to facilitate markerassisted breeding. Two of the most commonly used tools for analyzing complex traits are linkage analysis and association analysis or association mapping. That is, linkage disequilibrium (LD)-based methods are used to detect associations between phenotypic variations and genetic polymorphisms. Thus, association mapping analysis complements traditional linkage analysis in understanding the genetic basis of complex traits and offers several advantages over linkage analysis, such as much higher mapping resolution, large number of alleles, wide reference population, and less research time in establishing an association (Buckler and Thornsberry 2002;Darvasi and Shifman 2005;Flint-Garcia et al. 2003;Mackay 2001). Furthermore, Shi et al. (2011) stated that using breeding populations may be a more practical approach for crop variety development, considering that markers linked to major QTLs can immediately be utilized in markerassisted selection (MAS). Therefore, a thorough understanding of genetic diversity, population structure, and family relationships in a given panel is necessary for successful association analysis.
There are two methods to identify the genomic regions associated with important agricultural traits. One method is QTL mapping based on the results of crosses between two parents with linked and contrasting phenotypes and genotypes within segregated populations (Skot et al. 2005). The other method is association mapping using LD between molecular markers and agronomic traits of interest (Pritchard et al. 2000;Zhang et al. 2012). Many population structures have been used for crop QTL detection and mapping. For example, backcross populations, F 2 populations, doubled haploid populations, recombinant inbred line populations, and near-isogenic line populations have all proven useful in identifying and confirming QTLs using various molecular marker systems Farre et al. 2016;Park et al. 2013;Ramekar et al. 2018;Sa et al. 2015;Tanksley and Nelson 1996). However, in contrast with other major crops such as rice, wheat, and maize, QTL analysis using genetic maps of segregated populations is very difficult for Perilla crop because various molecular markers, such as simple sequence repeat (SSR) markers for each chromosome, have not yet been developed. In addition, more than 500 seeds of the F 2 population are produced in F 1 plant. Because of the large number of seeds, genetic analysis of the entire F 2 population is not feasible . Therefore, in this study, some lines of the F 2 population were selected based on leaf characteristics such as leaf color (green or purple) and leaf shape (nonwrinkled or wrinkled) (see Fig. 1) to identify morphological differences between the two cultivated types of Perilla crop using molecular markers.
Meanwhile, molecular marker-based techniques in genetic research have been used in many crop species to identify the genomic regions associated with important agricultural traits. Among the various DNA molecular marker systems, SSRs are codominant markers, which are considered to be one of the most suitable means for assessing genetic diversity, genetic relationships, population structure, QTLs, and association mapping because of their reliability, reproducibility, and discrimination (Park et al. 2009;Pejic et al. 1998;Sa et al. 2018;Vathana et al. 2019). Thus, for assessing population structure and association analysis, SSRs are widely used in many major crop species, such as rice (Pradhan et al. 2016), maize (Kim et al. 2017), wheat (Crossa et al. 2007), and soybean (Qin et al. 2016). However, this advantage of SSRs is only applicable to crop species with large numbers of generally freely accessible expressed sequence tag (EST) or cDNA sequences. In general, SSRs for crop species with poor genomic sequence information, such as most small or minor crop species, have not been developed (Zane et al. 2002;Squirrell et al. 2003;Park et al. 2009).
In Perilla crop, there has not been sufficient genetic research, even though SSR markers have recently been developed by several researchers, such as Kwon et al. (2005), Park et al. (2008) and Sa et al. (2018Sa et al. ( , 2019. In our previous study, we sequenced and assembled one cultivated type (PF98095) of P. frutescens var. frutescens using transcriptome sequencing by RNAseq (Tong et al. 2015), and we also obtained 15991 SSR loci . The transcriptome sequencing approach RNA-seq represents a powerful tool for transcriptional analysis, novel gene discovery, and the development of SSR markers for nonmodel crops (Mutz et al. 2012;Wang et al. 2015;Yang et al. 2018). In particular, SSR markers developed from RNA-seq can enable marker-assisted selection for crop breeding programs because these regions are coding sequences and may be close to or within functionally transcribed genes (Mutz et al. 2012;Yang et al. 2018). In this study, we successfully developed SSR primers from Perilla crop. These novel additional SSR markers can be used to analyze the genetic diversity, genetic relationships, population structure, and association analysis in Perilla species.

(b) (a)
Our study reports the results of the new Perilla SSR primers developed by the transcriptome sequencing approach RNA-seq and the results of association analysis in the F 2 population using these novel additional SSR primers and previously developed SSR primers. The results of this study are expected to provide useful information for future Perilla crop breeding programs.

Plant materials and morphological characteristics of the F 2 population
To develop the F 2 population of the Perilla crop, a cross was made between PF13-110 (female parent) and PF13-160 (male parent) ( Fig. 1; Table 1). The female parent, PF13-110, is a var. frutescens that has non-wrinkled leaves with a green leaf surface and reverse side, green stem, and a fragrance specific to the var. frutescens ( Fig. 1a; Table 1). The male parent, PF13-160, is a var. crispa that has wrinkled leaves with a purple leaf surface and reverse side, purple stem, and a fragrance specific to the var. crispa ( Fig. 1b; Table 1).
To detect SSR markers associated with morphological characteristics, 40 individuals of the F 2 population were selected based on leaf characteristics such as leaf color, leaf size, and leaf shape and plant characteristics such as pubescence degree, plant fragrance, and flower and stem colors (Tables 1, 2). Of the total individuals of the F 2 population used in this study, approximately 20 individuals generally showed typical var. frutescens characteristics in leaf color and leaf shape, and the other 20 individuals showed a typical shape of var. crispa ( Fig. 1; Table 2). To evaluate the morphological variation among the 40 individuals of the F 2 population and two parental lines of Perilla crop, each line was grown in pot (9 cm 9 9 cm) under normal greenhouse conditions from May to October at the College of Agriculture and Life Sciences, Kangwon National University, Chuncheon, Gangwon-do, Korea. We examined nine leaf-and plant-related traits, as shown in Tables 1 and 2, that were selected based on a previous report by Lee and Ohnishi (2001).

DNA extraction and SSR primer development
Total DNA was extracted from young leaf tissues following the Plant DNAzol reagent protocols (Gib-coBRL Inc., Grand Island, NY, USA). For construction of the transcriptome reference set in a previous study (Tong et al. 2015), de novo assembly of the PF98095 RNA-seq data was performed using Trinity software (http://TrinityRNASeq.sourceforge.net). The raw reads from next-generation sequencing (NGS) with a Phred quality score of at least 20 and a read length of at least 50 bp of HiSeq 2000 data were   (Bradbury et al. 2007) was used to evaluate marker-trait associations using a general linear model (GLM). The number of permutation runs was set to 10,000 to obtain a marker significance value of P B 0.05. Power Marker version 3.25 (Liu and Muse 2005) was applied to obtain information on the number of alleles, major allele frequency (MAF), genetic diversity (GD), and polymorphic information content (PIC). GD was calculated for each pair of accessions using the Dice similarity index (Dice 1945). To examine the clustering pattern of all individuals of the F 2 population by construction of a phylogenetic tree, Nei's distance was calculated and used for the construction of unrooted phylogeny using a neighbor-joining (NJ) method, bootstrapping of the data (10,000 permutations) was implemented in POWERMARKER, and MEGA X was used to visualize the tree (Nei 1973;Kumar et al. 2018).

SSR identification and polymorphisms
In our study, we surveyed the 148 newly developed SSR primer sets using nine Perilla accessions ( Table 2). Meanwhile, we surveyed the 39 newly developed SSR primer sets between the two parental lines of the F 2 population of the Perilla crop. Based on the results, we selected 12 SSR primer sets that exhibited good amplification patterns and polymorphisms between the two parental lines of the F 2 population of the Perilla crop (Tables 3, 4). In addition, we surveyed a total of 105 SSR primer sets that were developed in previous research. Among these, we selected 25 SSR primer sets that showed polymorphism between the two parental lines of the F 2 population (Table 4). Therefore, in our study, a total of 37 SSR primer sets, including 12 new SSR primer sets and 25 previous SSR primer sets, were used to measure polymorphisms in terms of GD, PIC, MAF, and separation patterns of allele bands (SPAB) among the 40 individuals of the F 2 population of the Perilla crop (Table 4). In the results, the GD ranged from 0.185 (KNUPE95) to 0.708 (KNUPE102), with an average  ). Therefore, these SSR primer sets are considered to deviate from the expected Mendelian segregation ratio of 1:2:1 (or AA:AB:BB) in the F 2 population because more than 50% of the allele bands amplified in the F 2 population appeared to be biased towards one of the two parents (A or B). The remaining SSR primer sets showed a tendency towards a Mendelian segregation ratio of 1:2:1 (or AA:AB:BB) in the 40 individuals of the F 2 population, although this does not represent the entire F 2 population. Meanwhile, among the SSR primer sets used in the analysis, 21 SSR primer sets showed a null band pattern in the 40 individuals of the F 2 population (Table 4).
In addition, a phylogenetic tree was constructed using a total of 37 selected SSR markers to elucidate the phylogenetic relationship between the 40 individuals of the F 2 population and their two parental lines (Fig. 2). The 40 individuals of the F 2 population and two parental lines were clustered into two major groups. The first group (Group I) included 18 individuals of the F 2 population, including Parent B (PF13-160). The second group (Group II) included 22 individuals of the F 2 population, including Parent A (PF13-110). Although there were some exceptions, the results showed that Group I mainly contained many individuals with similar characteristics to Parent B and that Group II included many individuals with similar characteristics to Parent A (Table 2; Fig. 2).

Phenotypic variation and association analysis of leaf-and plant-related traits
The distribution of leaf-and plant-related traits in the 40 individuals of the F 2 population was examined (Table 5). For leaf size (QL1), 12 individuals showed large, 14 individuals showed medium, and 14 individuals showed small. For color of leaf surface (QL2), 28 individuals showed green, 6 individuals showed mixed green/purple, and 6 individuals showed purple.
Meanwhile, to select SSR markers associated with leaf-and plant-related characteristics in the 40 individuals of the F 2 population, genotypes of the 37 SSR markers and the nine leaf-and plant-related traits were used to confirm significant marker-trait associations (SMTAs) using TASSEL software (Table 7). In the results, we detected 24 SMTAs involving 21 SSR markers associated with nine leaf-and plant-related traits using GLM at a significance level of P B 0.05 (Table 7. Among the 12 SSR markers for leaf-related traits, KWPF25 and KNUPF108 were associated with QL2, KNUPF11 was associated with QL3, KNUPF74 and KNUPF89 were associated with QL5, and seven SSR markers (KNUPF21, KNUPF49, KNUPF60, KNUPF81, KNUPF103, KNUPF109, GBPEM111) were associated with QL8. Also 11 SSR markers were associated with plant-related traits. Among them, two SSR markers (KNUPF15 and KNUPF21) were associated with QL4, KNUPE11 was associated with QL6, four SSR markers (KNUPF11, KNUPF54, KNUPF82, and KNUPF88) were associated with QL7, and five  Table 5 Distribution of morphological variation in 40 individuals of the F 2 population developed from cross between Perilla frutescens var. frutescens and var. crispa

Discussion
Genetic variation between individuals within a population or between populations of crop species because of either or both genetic and environmental influences can be easily assessed using various molecular markers. Among various molecular markers, SSR markers in particular have many advantages compared with other marker systems. One advantage is high reproducibility, which is most important in genetic analysis. A second advantage of the SSR marker system is the polymorphic genetic information contents, and the hypervariable nature of SSRs produces very high allelic variations even among very closely related varieties. A third advantage relates to the codominant nature of SSR polymorphisms because the codominant nature of SSRs is suitable for genetic analysis in segregation of F 2 populations or pedigree analysis in hybrids (Park et al. 2009;Pejic et al. 1998;Sa et al. 2018).
In plants, the presence of SSRs was first demonstrated by the hybridization of poly (G-T) and poly (A- G) oligonucleotide probes on phage libraries of tropical tree genomes (Condit and Hubbell 1991). A search of published DNA sequences revealed that SSRs are also abundant in diverse plant genomes (Morgante and Olivieri 1993;Wang et al. 1994). Therefore, SSRs have become the preferred molecular marker system for the analysis of plant genetics and ecology. However, SSR markers can only be applied to crop species with large amounts of EST or cDNA sequences that are freely accessible to the public. Recently, the transcriptome sequencing approach RNAseq has become a powerful tool for novel gene discovery and the development of molecular markers for nonmodel crop, such as Perilla crop. In a previous study of Perilla crop we sequenced and assembled one cultivated type (PF98095) of P. frutescens var. frutescens using transcriptome sequencing by RNAseq (Tong et al. 2015). This information will help develop SSR primers for Perilla crop.
As explained in the Introduction, Sa et al. (2018) detected a total of 15991 SSR loci based on the SSR flanking sequences via de novo assembly of the PF98095 RNA-seq data (Tong et al. 2015), which were classified based on the number of repeating units. Most SSRs were dinucleotide SSRs (9910, or 62.0%), followed by trinucleotide SSRs (552, or 35.3 %) and tetranucleotide SSRs (429) . In this Perilla crop analysis dinucleotide repeats and trinucleotide repeats together accounted for 97.3% of the eSSR repeats. This finding is important for developing a large number of effective eSSR primers in Perilla crop because the markers associated with smaller SSR motif lengths (di-and tri-) are more variable than those associated with other motif lengths (Ellegren 2004). Our results also showed that dinucleotide repeats were more abundant than other repeat units in Perilla crop. While an almost equal distribution of di-and trinucleotide repeats was reported in ramie (Liu et al. 2013), a high abundance of trinucleotide repeats was reported in flax (Cloutier et al. 2009), jute (Saha et al. 2017), and kenaf ESTs (Zhang et al. 2015). In this present study, the dinucleotide SSRs such as KNUPF87, KNUPF97, KNUPF103, KNUPF105, KNUPF110, and KNUPF116 showed comparatively more allelic bands and genetic diversity values than the tri-and tetranucleotide SSRs in the Perilla accessions. However, in many crop species the repeat motifs may sometimes show different amplification tendencies depending on the crop-specific sequences of repeat motifs. Therefore, a description of the allelic band and genetic diversity of SSRs based on their repeat motifs can be different because of the genomic sequence conditions of each crop species. Although we analyzed only nine Perilla accessions, the allele number, GD, PIC, and MAF values for 39 new Perilla EST-SSR markers were found to determine the unique genetic profiles of individual genotypes of the Perilla crop. Therefore, these new SSR primer sets may be useful in the future for studying GD for Perilla germplasm resources by association mapping and designating important genes/QTLs for future genetic and breeding programs in Perilla crop.
Analysis of genetic diversity and relationships and association mapping of plant breeding materials will help develop new molecular markers associated with target morphological traits or useful varieties and lines in Perilla crop breeding programs. In our study, we developed an F 2 population from a cross between PF13-110 (female parent, cultivated var. frutescens) and PF13-160 (male parent, weedy var. crispa) to develop molecular markers for traits showing morphological differences between the two cultivated types of var. frutescens and var. crispa through the 40 individuals of the F 2 population of the Perilla crop ( Fig. 1; Tables 1 and 2). According to our results, the 40 individuals of the F 2 population showed discontinuous variations within the range of morphological characteristics of their parental lines, as shown in Tables 2 and 5. That is, some morphological traits investigated in our study showed a Mendelian segregation ratio of 1:2:1 (or AA:AB:BB), while other traits showed different separation patterns in the 40 individuals of the F 2 population. For example, in the case of QL1, of the 40 lines of the F 2 population, 12 lines were large, 14 lines were medium, and 14 lines were small; whereas in the case of QL2, 28 lines were green, 6 lines were mixed green/purple, and 6 lines were purple. These results indicate that the leaf-related traits in the F 2 generation of Perilla crop exhibit different genetic segregation patterns because of incomplete dominance or interaction of one or more alleles. Award and Mohamed (2017) reported that the flower color of two parental types ($ violet, # white) and the F 1 hybrid type (light violet corolla) of Catharanthus roseus (L.) plant showed a Mendelian segregation ratio of 1:2:1 in the F 2 generation; therefore this flower color was related to monogenic control and incomplete dominance. Meanwhile, Koezuka et al. (1985) reported that the segregation ratio of red and green plants in the F 2 population of Perilla crop was 3:1 for red and green leaf color, which is controlled by a single gene, and was 15:1 for red and green stem color, which is controlled by two independent genes. However, unfortunately, as mentioned in the Introduction, the F 2 population lines used in this study did not represent the entire population, so it was not possible to predict the exact segregation ratio of the F 2 population.
In our study, we also performed a correlation analysis for nine leaf-and plant-related traits in the 40 individuals of the F 2 population (Table 6). Among the morphological traits examined in our study, leaf and plant color-related traits, such as QL2, QL3, QL6, and QL7, and leaf shape-related traits, such as QL5 and QL8, were closely related to each other. Therefore, these traits are expected to be useful traits for distinguishing or selecting the two parental lines of var. frutescens and var. crispa and their F 2 population. Moreover, to better understand the genetic relationships among the 40 individuals of the F 2 population, we analyzed the phylogenetic relationships of the 40 individuals of the F 2 population and their two parental lines using 37 SSR markers. In the phylogenetic tree analysis, the 40 individuals of the F 2 population and two parental lines were clustered into two major groups. Except for exceptional individuals, Group I contained many individuals with similar characteristics to Parent B, while Group II contained many individuals with similar characteristics to Parent A (Table 2; Fig. 2). Therefore, these separated individuals and leaf-and plant-related traits of the F 2 population are considered to be useful genetic materials for finding molecular markers associated with the leaf and plant morphological characteristics of Perilla crop through association mapping analysis.
Association mapping has been proposed as a selection method to identify loci involved in the inheritance of complex traits (Risch and Merikangas 1996). In our study, we developed an F 2 population for association mapping analysis to find molecular markers associated with leaf-and plant-related traits of Perilla crop. Also, to select the SSR markers associated with leaf-and plant-related traits of the F 2 population, we analyzed marker-trait associations (SMTAs) between 37 SSR markers and 9 leaf-and plant-related characteristics in 40 individuals of the F 2 population using TASSEL software. From the results, we identified 12 SSR markers associated with leafrelated traits and 11 SSR markers associated with plant-related traits (Table 7). These 12 SSR markers (KWPF25, KNUPF11, KNUPF21, KNUPF49,  KNUPF60, KNUPF74, KNUPF81, KNUPF89,  KNUPF103, KNUPF108, KNUPF109, and GBPEM111) were associated with the leaf-related traits QL2, QL3, QL5, and QL8. The 11 SSR markers (KNUPF5, KNUPF11, KNUPF15, KNUPF21, KNUPF29, KNUPF39, KNUPF54, KNUPF61, KNUPF82, KNUPF88, and GBPEM111) were associated with the plant-related traits QL4, QL6, QL7, and QL9 (Table 7). Our results showed that, with the exception of QL1, most leaf-and plant-related traits were associated with 1 to 7 SSR markers, depending on the leaf-and plant-related characteristics. Additionally, four SSR markers (KNUPF11, KNUPF21, GBPEM111) were associated with leaf-or plantrelated characteristics. For example, KNUPE11 was found to be associated with QL3, QL6, and QL7 traits; GBPEM111 was associated with QL8 and QL9 traits; and KNUPE21 was associated with QL4 and QL8 traits (Table 7).
In a previous study by Lim et al. (2021), bulk segregant analysis (BSA) was performed to identify SSR markers linked to leaf-and seed-related traits in Perilla crop. They reported that four SSR markers (KNUPF15, KNUPF21, KNUPF29, and KNUPF60) were associated with the color of the leaf surface, four SSR markers (KNUPF11, KNUPF15, KNUPF21, and KNUPF60) were associated with the color of the leaf reverse side, and three SSR markers (KNUPF25, KNUPF61, and GBPEM111) were associated with the depth of leaf teeth. In addition, five SSR markers (KNUPF11, KNUPF12, KNUPF16, KNUPF29, and KNUPF42) were associated with seed-related traits. According to our results, among the SSR markers associated with the leaf-and plant-related traits, eight SSR markers (KNUPE11, KNUPF15, KNUPF21, KWPE25, KNUPF29, KNUPE60, KNUPE61, GBPEM111) appeared to overlap with leaf-related traits previously reported by Lim et al. (2021). Therefore, these SSR markers are thought to be useful molecular markers for selecting leaf-and plant-related traits in Perilla crop. Additionally, these SSR markers are thought to be useful molecular markers for distinguishing between the two cultivated types of Perilla crop. However, in Perilla crop, it is difficult to compare genetic characteristics because genomic information on SSR markers associated with leaf-and plant-related characteristics is still lacking.
In the future, if studies such as genome analysis for each chromosome are actively conducted in Perilla crop, it will be possible to analyze genomic information of specific Perilla SSR markers related to leafand plant-related characteristics. As explained in the Introduction, many taxonomic studies have been conducted to distinguish between the two cultivated and weedy types of Perilla crop using morphological characteristics and molecular markers, but they failed to distinguish the two cultivated types of Perilla crop because of the existence of intermediate types, such as weedy plants. Therefore, the SSR markers associated with leaf-and plant-related characteristics selected in this study are expected to provide useful information for distinguishing the two cultivated types of Perilla crop and their weedy types in the future.

Conclusions
In this study, we developed 39 new SSR primer sets by the transcriptome sequencing approach RNA-sEq. In total, 130 alleles were detected segregating in nine Perilla accessions with an average of 3.3 alleles per locus, ranging from 125 to 360 bp. The number of alleles per locus ranged from two to six. We also developed an F 2 population from a cross between PF13-110 (female parent, cultivated var. frutescens) and PF13-160 (male parent, weedy var. crispa) to develop molecular markers for traits showing morphological differences between the two cultivated types of Perilla crop, namely var. frutescens and var. crispa. To detect SSR markers associated with morphological characteristics of Perilla crop, 40 individuals from the F 2 population of Perilla were selected for association analysis based on their leaf-and plantrelated characteristics. In an association analysis of 37 SSR markers and 9 leaf-and plant-related traits in the 40 individuals of the F 2 population, we identified 12 SSR markers associated with leaf-related traits and 11 SSR markers associated with plant-related traits. Among the SSR markers associated with the leafand plant-related traits, eight SSR markers (KNUPE11, KNUPF15, KNUPF21, KWPE25, KNUPF29, KNUPE60, KNUPE61, GBPEM111) appeared to overlap with leaf-related traits previously reported by Lim et al. (2021). Therefore, the SSR markers associated with leaf-and plant-related characteristics selected in this study are expected to provide useful information for distinguishing the two cultivated types of Perilla crop. In addition, the new Perilla SSR primers described in this study could be helpful in identifying genetic diversity and genetic mapping, designating important genes/QTLs for Perilla crop breeding programs, and allowing Perilla breeders to improve leaf and plant quality through MAS breeding programs.