Microsatellites, or simple sequence repeats (SSRs), are tandem repeats of mono- to hexa-nucleotide sequence motifs that generally vary in length between five and 40 repeats. They are codominant, abundant, multiallelic and locus-specific markers and are found at high frequency in eukaryotic genomes [1, 2]. Microsatellites are powerful markers for population genetic studies on breeding systems, genetic diversity, and conservation genetics [3, 4]. However, traditional approaches used to identify microsatellite markers are usually time-consuming, labor-intensive, and costly processes [4]. Recently, next-generation sequencing (NGS) has been accepted as a rapid and cost-effective solution for the identification of microsatellite markers in nonmodel plants [5, 6]. Microsatellites can be identified in both genomic and transcriptomic sequences with contrasting features. For example, genomic SSRs have an uncertainty associated with protein-coding genes, while transcriptomic SSRs are potentially linked with protein-coding genes and their untranslated regions (UTRs) [4]. Genomic SSRs are highly polymorphic and are randomly dispersed throughout the genome, while transcriptomic SSRs exhibit relatively low polymorphism and occur infrequently [4, 7]. However, transcriptomic SSRs facilitate the understanding of their association with functional genes or phenotypic variation [8]. Transcriptomic SSRs are transferable among closely related species, whereas genomic SSRs have relatively low transferability [4]. Furthermore, transcriptomic SSRs may play an important role in adaptive evolution by generating genetic variation [9].

The angiosperm genus Lychnis (Caryophyllaceae) is well studied with respect to the genetic effects of habitat fragmentation on population structure [10,11,12]. Many Lychnis species are drought-tolerant perennials, but some of them require water to survive. Previous studies have shown that drought stress, mating history and nutrient availability influence inbreeding depression in Lychnis populations [13, 14]. This genus exhibits diversity in their sexual and mating systems, breeding system and host-pathogen dynamics, sex chromosome evolution, genomic conflict and speciation, and biological invasions [15]. Lychnis kiusiana Makino is an endangered perennial herb native to wetland areas in Korea and Japan [16]. The loss of wet meadow habitats has resulted in decreases in the size and number of populations of this species [17]. Fragmentation of habitats may influence the breeding system and may lead to a reduction in genetic variation, resulting in diminished population capacity to adapt to environmental changes [18,19,20,21]. Information on genetic diversity is important for planning the management of L. kiusiana. Despite its conservational and evolutionary significance, population genetic resources are lacking for L. kiusiana. Only five polymorphic genomic SSR markers have been identified using traditional approaches, and an applied genetic diversity study has been conducted in Japanese populations of the species [17]. However, enriched genomic resources and genetic markers are needed to interpret genetic variation within populations and to investigate the genetic diversity of L. kiusiana. Transcriptomic sequencing has been applied for mining microsatellite markers in related species, i.e., Silene vulgaris [22], and the results indicated that transcriptomic SSRs provide potential for genomic and population genetic approaches.

In this study, we developed microsatellites from the L. kiusiana transcriptome and characterized their frequency, distribution, and function, which promotes an understanding of microsatellite evolution in the L. kiusiana transcriptome. We estimated nonsynonymous (dN) and synonymous (dS) substitution rates of open reading frames (ORFs) that contained at least one single-nucleotide polymorphism (SNP). In addition, we designed primers for amplifying microsatellite loci and validated the availability of selected primers. Our results offer a valuable resource for studies of population genetics in L. kiusiana as well as three other Lychnis species.


Characteristics of microsatellites in the Lychnis kiusiana transcriptome

Using the Illumina HiSeq 2000 sequencing technology tool, we assembled 67,498,600 reads into 91,900 contigs with an average length of 800 bp. A total of 11,403 microsatellite repeat motifs were identified from 9563 contigs (10.4%; Additional file 1: Table S1). Of the total 11,403 SSRs, tri-nucleotide microsatellites were the most abundant motif (42.2%), followed by mono- (35.3%), di- (13.7%), hexa- (9.2%), penta- (3.6%), and tetra-nucleotide (1.7%) types (Additional file 2: Figure S1). A/T motifs (97.7%) were the most abundant in mono-nucleotide repeats (Additional file 2: Figure S1). With respect to the di-nucleotide motif, AG/CT was the most abundant type, with a total of 800 (51.4%), while there was only one CG/GC. Ten tri-nucleotide motif types were identified with AAT/ATT accounting for approximately 21.4%, while CCG/CGG only accounted for 1.7% (Additional file 2: Figure S1). The hexa-nucleotide motif accounted for the highest number (77) across all types (Additional file 2: Figure S1).

The average length of the microsatellites from the L. kiusiana transcriptome was 19.75 bp (Additional file 2: Figure S2A). The length variation of microsatellites was significantly affected by the repeat motif size (Kruskal-Wallis rank sum test, p < 1 × 10− 15; Additional file 2: Figure S2B). The length differences between each motif size class were statistically significant (pairwise Wilcoxon rank-sum test, p < 1 × 10− 15 after Bonferroni correction; Additional file 2: Figure S2B), except for two comparisons between mono- and di-nucleotide and between tetra- and penta-nucleotide motifs. The mono-nucleotide motifs had the shortest average length (16.7 bp), while hexa-nucleotide motifs were the longest, with an average of 27.4 bp. The longest microsatellite identified was 96 bp, which was composed of a 16-fold repetition of a hexa-nucleotide motif.

Microsatellite distribution in the Lychnis kiusiana transcriptome

The distribution of microsatellites in the 5′ UTR, CDS region, and 3′ UTR was investigated. Of 11,403 microsatellites, 2115, 3543 and 1818 were located in the 5′ UTR, CDS region, and 3′ UTR, respectively (Additional file 1: Table S2). The remaining 3927 microsatellites were excluded from analyses because the transcripts lacked information to determine the CDS region. The frequency of microsatellites in the CDS region was similar to that in the UTRs (Additional file 1: Table S2). However, the frequencies of the motif size classes were significantly affected by the location (Kruskal-Wallis rank sum test, p < 1 × 10− 15; Fig. 1). Microsatellites located in the CDS regions were mostly tri-nucleotide motifs (72.1%). Mono- and di-nucleotide microsatellites dominated in the UTRs, showing that the proportions of mono- (mean 39.8%) and di-nucleotide motifs (mean 19.9%) in the UTRs were much higher than those in the CDS regions (Fig. 1). The mean lengths of microsatellites differed significantly between the CDS regions and each UTR (Kruskal-Wallis rank sum test, p < 1 × 10− 15; Additional file 2: Figure S3A). The mean microsatellite lengths of only two motif size classes were significantly affected by location (Kruskal-Wallis rank sum test, mono-: p < 1 × 10− 13, di-: p = 0.009301; Additional file 2: Figure S3B).

Fig. 1
figure 1

Frequencies of SSRs in different genic regions

Functional annotation of genes containing microsatellites

Gene Ontology (GO) assignment was used to classify the transcripts according to their function. Of the 9563 contigs in L. kiusiana, 6560 (68.6%) had blast hits against “land plants (taxa: 3193)” of the nonredundant (nr) database. In total, 4510 (47.2%) microsatellite-containing transcripts had GO annotations and were categorized into three functional groups and 50 subgroups (Fig. 2). Cellular (2401) and metabolic processes (2429) were the top two subgroups that involved the most genes in the “biological process” group. In the “cellular component” group, the cell (2259) and cell parts (2246) were the top two subgroups that involved the most genes. Two subgroups, catalytic activity (2241) and binding (2239), involved the most genes among the “molecular function” group. A suite of transcripts was annotated as transcription factors (TFs) that contain typical DNA binding motifs, such as AP2/ERF, basic leucine zipper (bZIP), MADS-box, MYB, and WRKY (Additional file 1: Table S3). The functions of microsatellite-containing transcripts were further surveyed by KEGG pathway analysis. The results showed that the transcripts were involved in a total of 124 pathways (Additional file 1: Table S4). The KEGG pathways, including metabolism (92.9%), organismal systems (5.0%), genetic information processing (1.1%), and environmental information processing (1.0%), were grouped into 16 functional categories (Fig. 3). The most represented pathways included purine, pyrimidine and thiamine metabolism, biosynthesis of antibiotics, aminobenzoate degradation, the T-cell receptor signaling pathway, Th1 and Th2 cell differentiation, and starch and sucrose metabolism (Additional file 1: Table S4).

Fig. 2
figure 2

Classification of genes containing microsatellite loci based on the Gene Ontology (GO) annotation

Fig. 3
figure 3

KEGG pathways involved the genes containing microsatellites

Identification of selection signatures in transcriptomic SSR markers

To identify transcriptomic SSR markers under selection, we calculated SNPs on transcripts and investigated their effect on nucleotide substitution rates. A total of 67,498,600 reads were mapped onto the 9563 transcripts that contained microsatellites, generating 12,486 SNPs after quality control and filtration. The proportions of transition substitutions were 30.5% for A/G and 29.9% for C/T, compared with smaller proportions of transversion for A/C (9.5%), A/T (13.1%), C/G (7.3%), and G/T (9.7%). Among all SNPs detected, 3087 were in the putative 2389 ORF regions, of which 1943 were nonsynonymous (62.9%) and 1144 were synonymous (37.1%). We estimated nonsynonymous (dN) and synonymous (dS) substitution rates of the ORFs that contained at least one SNP. The results showed that the dN/dS ratios of 39 ORFs were greater than one, indicating that the loci are putatively under positive selection (Fig. 4). Investigation of these ORFs may represent genes with interesting evolutionary histories that have led to high levels of accumulated polymorphism. Functional annotations for 39 ORFs showed that some of the ORFs included genes encoding Cullins, ABC transporters, serine/threonine-protein kinase SRK2E, and synaptotagmin-5 (Additional file 1: Table S5).

Fig. 4
figure 4

Correlation of synonymous and nonsynonymous substitution rates of SNPs. The solid red line indicates the dN/dS ratio is equal to one. The dashed line represents the regression, which was analyzed using dN and dS for all unigenes. The red circles represent unigenes with dN/dS ratios > 1

Validation of transcriptomic SSR markers

The primers of 6911 microsatellites were designed based on 9563 transcripts containing microsatellites. Apart from transcriptomic SSR markers under selection, a total of 50 pairs of primers were selected (except mononucleotide repeats). Ten primer pairs gave no amplicon, but 40 pairs were consistently and successfully amplified from 51 individuals of the Minamioguni (MG, Kumamoto, Japan) population. Of the 40 primer pairs, 38 produced an amplicon of the expected size, whereas the others produced amplicons larger than the expected size. Twenty-five microsatellite loci were polymorphic and were successfully genotyped in the 51 individuals of MG, generating a total of 174 alleles. The average number of alleles per SSR locus was 6.96 (Table 1). The estimated null allele frequency for LKI20 was greater than 0.1 (Table 1), and MICRO-CHECKER also indicated a higher frequency of null alleles at the same locus. No significant deviation from Hardy-Weinberg equilibrium (HWE) was found except for one (LKI20), following the correction based on the false discovery rate (p < 0.05). A linkage disequilibrium (LD) test revealed that most of the loci were in linkage equilibrium except three pairs of loci (LKI05 and LKI17, LKI16 and LKI17, and LKI20 and LKI34) (Additional file 2: Figure S4). Observed (HO) and expected (HE) heterozygosities ranged from 0.1373 to 0.9020 and 0.1314 to 0.8270, respectively, and locus-specific FIS estimates ranged from − 0.116 (LKI29) to 0.290 (LKI20). The positive FIS value of LKI20 was significantly higher than zero (Wilcoxon signed-rank test; p < 0.05). The polymorphism information content (PIC) of 25 SSR markers ranged from 0.126 to 0.798. At the population level, the MG population exhibited high genetic diversity (HO = 0.6455, HE = 0.6617) and a low inbreeding coefficient (FIS = 0.023). The mean value of the Garza-Williamson (G-W) index for the MG population was 0.8278. Twenty-one of the identified microsatellite markers had GO annotations (Additional file 1: Table S6). In particular, LKI09 which exhibited a high level of polymorphism (Table 1) was associated with WRKY transcription factor 57 (WRKY57) (Fig. 5).

Table 1 Summary statistics for 25 polymorphic microsatellite loci developed for Lychnis kiusiana. Number of alleles (NA), observed (HO) and expected (HE) heterozygosities, inbreeding coefficient (FIS), Hardy-Weinberg equilibrium (HWE), and polymorphism information content (PIC). Significant values after false discovery rate correction are indicated with asterisks (* p < 0.05, ** p < 0.01)
Fig. 5
figure 5

The LKI09 transcriptomic SSR. a. Schematic diagram of the LKI09 transcriptomic SSR surrounding the WRKY57 gene from L. kiusiana. Dark and bright green arrows indicate the forward primer and reverse primer, respectively. The blue box indicates a microsatellite. The red box indicates a conserved domain (WRKY). Purple and orange boxes indicate the well-known WRKYGQK motif and a zinc-binding site, respectively [29]. b. DNA and amino acid sequences of the nuclear-encoded WRKY conserved domain, including the UTRs. Colored boxes indicate primers, microsatellites, conserved domains, motifs and zinc-binding sites corresponding to A. Asterisks indicate stop codons

To examine cross-species transferability, 25 newly developed microsatellite markers were tested on the other Lychnis species (L. cognate, L. fulgens, and L. wilfordii). Eleven markers were successfully amplified in all three Lychnis species (44%; Table 2). Twenty and 16 loci were amplified in L. fulgens and L. cognate, respectively, while only 14 loci were amplified in L. wilfordii (Table 2).

Table 2 Amplification success of 25 microsatellites developed for Lychnis kiusiana across three other Lychnis species. Gray shading indicates a locus that was successfully amplified in all three Lychnis species


Deep transcriptome sequencing technologies are rapid and cost-effective tools to characterize gene content and identify polymorphic markers in nonmodel plants [6]. In this study, we generated a leaf transcriptome from the angiosperm L. kiusiana and characterized microsatellites in the transcriptome. The results showed that microsatellites were present in a small proportion of the transcripts (10.4%), which is higher than the estimation that 2–5% transcripts contain microsatellites [23]. These transcripts are involved in a wide range of potential functions, such as cellular and metabolic processes, cell and cell parts, and catalytic activity and binding (Fig. 2). The transcriptomic SSR markers provide a valuable resource for the development of genetic markers in L. kiusiana and offer a chance to classify the functions of these microsatellite-containing genes. In addition, a moderate level of transferability suggests that the microsatellite loci may be useful for studies of the other Lychnis species, although screening of microsatellite markers will be required to test polymorphisms.

Microsatellite distributions in genes, including coding regions, UTRs, and introns, are nonrandom and strongly biased [24]. Our results showed that the repeat motif size classes were significantly affected by location in the CDS regions or UTRs (Fig. 1). Of the six motif size classes, tri- and hexa- nucleotides were found in CDS regions with a high frequency (87.4%) because these types of microsatellites are less likely to cause frameshift mutations. In L. kiusiana, 5′-UTRs and 3′-UTRs contain more mono- and di-nucleotide motifs than coding regions. The 5′-UTRs contained more tri-nucleotide motifs than the 3′-UTRs in L. kiusiana (30.4% vs 24.9%). This is because microsatellite variations in 5′-UTRs influence gene expression and lead to protein adaptation [24]. Microsatellite length also reflects the effect of evolution and selection on microsatellite loci development. The Lychnis transcriptomic SSRs in UTRs were much longer than those in CDS regions (Additional file 2: Figure S3), reflecting higher evolutionary constraints on the microsatellites in the CDS regions than in UTRs.

Microsatellite repeat motif variation can influence gene regulation, transcription, translation, and protein function [9]. We found that many microsatellites in the L. kiusiana leaf transcriptome were linked to the genes involved in plant response to light intensity, salt stress, temperature stimulus, and nutrient and water deprivation. Some of these transcripts, such as aldehyde oxidase and cytochrome P450, are related to resistance to environmental stresses and insecticides [25, 26]. Lychnis kiusiana inhabits wetlands in mountain areas of Korea and Japan [16], and wetland plants are exposed to environmental factors such as salinity, soil anaerobiosis, low nutrient conditions, sediment deficiency, and fluctuating water regimes [27]. Previous studies have shown that inbreeding depression in Lychnis populations is affected by drought stress, mating history and nutrient availability [13, 14]. Thus, microsatellites located in these genes may be crucial for the accumulation of adaptive genetic variations that enable L. kiusiana to thrive in wetlands of mountain areas.

Transcription factors (TFs) play a crucial role in controlling cellular processes, such as plant growth and development pathways [28]. The Arabidopsis and rice genomes contain some transcripts harboring microsatellites that are related to TFs [29]. In the L. kiusiana leaf transcriptome, we found that many transcripts harboring microsatellites have transcription factor activity (Additional file 1: Table S3). Interestingly, among the 25 transcriptomic SSR markers, high levels of microsatellite polymorphisms (15 alleles) were detected in the LKI09 marker linked to a gene encoding probable WRKY57 (Fig. 5 and Table 1). WRKY TFs are key regulators of many processes in plants, including the responses to biotic (e.g., pathogen) and abiotic (e.g., drought and cold) stresses, leaf senescence and seed dormancy, germination and developmental processes [30]. In addition to WRKY TFs, AP2/ERF or MYB TFs also regulate various biological processes, including development, defense, and biotic and abiotic stress responses [31,32,33,34]. Many transcripts associated with the AP2/ERF or MYB TF family were identified in the L. kiusiana leaf transcriptome (Additional file 1: Table S3). These microsatellites related to TFs may function as an important “tuning knobs” for the expression of genes [35, 36]. Examination of microsatellite polymorphisms in Lychnis transcripts linked to TFs requires further study to provide valuable insights into understanding environmental stresses.

For genetic diversity study, including structure analysis, it is critical to identify loci affected by selection in order to exclude them. We applied outlier tests based on detecting SNP loci under selection. Examination of the rate of variation in the ORFs that contained at least one SNP revealed the acceleration of the evolution of some genes, probably to adapt to the extreme environmental stress in wetlands of mountain areas. For example, the dN/dS ratios of two transcripts, Cullins and ABC transporters, were higher than one (Additional file 1: Table S5). Cullins are a family of hydrophobic proteins that provide a scaffold for ubiquitin ligases (E3). E3 is a member of a ubiquitination system that operates in conjunction with an E1 (ubiquitin activating enzyme; UBA) and an E2 (ubiquitin conjugating enzyme; UBC) and functions in regulating abiotic stress (e.g., drought, salinity, cold and nutrient deprivation) responses [37]. In plants, ATP-binding cassette transporters (ABC transporters) are involved in essential aspects of a terrestrial plant’s lifestyle [38]. ABC transporters also play an important role in organ growth, plant nutrition, plant development, response to abiotic stress, and the interaction of the plant with its environment [39].

The genetic variability of microsatellites is extensively exploited in evolutionary studies of a wide variety of plant species [40]. Genetic diversity and fitness play an important role in species conservation, especially in rare plants and small populations [41, 42]. Lychnis kiusiana is an endangered perennial herb native to wetland areas, and the sizes and numbers of populations of this species are decreasing [17]. A previous study showed high genetic diversity (mean HE = 0.791 and mean NA = 12.0) within the seven L. kiusiana populations based on five genomic SSR markers [17]. Although these novel genomic SSR markers are useful for investigating genetic diversity and molecular-assisted breeding, these markers are still insufficient for genetic population study. Herein, we identified applicable microsatellites from the L. kiusiana leaf transcriptome, and 25 transcriptomic SSR markers have been proven to be efficient genetic markers. Our results showed that the genetic diversity of the MG population with the 25 transcriptomic SSR markers was higher (NA = 6.96 and HE = 0.6617), although the five genomic SSRs revealed higher genetic variations (NA = 14.2 and HE = 0.817) than the transcriptomic SSR in the MG population. This is because genomic SSR markers generally have a higher level of polymorphism compared with transcriptomic SSR markers [7]. Compared with other transcriptomic SSR markers [43,44,45,46], the developed L. kiusiana transcriptomic SSR markers have a high level of genetic diversity. Full understanding of the genetic diversity and structure of L. kiusiana requires evaluation of genetic variation among the other Korean and Japanese populations. SSR neutrality tests in the populations of L. kiusiana are also needed to identify either neutral or outlier loci among the 25 newly developed markers.

Given the high transferability of transcriptomic SSR markers, the microsatellites identified from L. kiusiana will have wide application in other Lychnis species. We applied 25 newly developed transcriptomic SSR markers to three species of the genus Lychnis to evaluate the transferability of these markers. The results showed that the transferability ratios were 56% in L. wilfordii, 64% in L. cognate, and 80% in L. fulgens, consistent with the estimation (46.8–100%) for a genus [7]. The high transferability of L. kiusiana transcriptomic SSRs to L. fulgens may be due to a high similarity of the sequences flanking the SSR between two species. In particular, the 14 amplified SSR markers in L. wilfordii will be a valuable source for its population genetic analyses because L. wilfordii is an endangered species in Korea.


In this study, we characterized a large number of transcriptome-derived microsatellites from Lychnis kiusiana, and functional annotation of microsatellite-containing transcripts provides new insights into the evolution of microsatellites. Several microsatellites in the L. kiusiana leaf transcriptome were linked to the genes involved in plant response to biotic and abiotic stresses. The microsatellites located in these genes may be crucial for L. kiusiana to evaluate adaptive genetic variations to adapt to wetlands in mountain areas. Our results showed that the identified microsatellite loci that have high allele numbers and heterozygosity could provide genetic markers in L. kiusiana populations. Furthermore, many microsatellites were transferable across the other species of the genus.


Sample collection, RNA extraction and sequencing

Leaf samples of Lychnis kiusiana were collected from a single individual in the Korea National Arboretum, South Korea. Total RNA was extracted from fresh leaf tissue (100 mg) using the methods of Ghawana et al. [47] and was treated with DNase I (Invitrogen, California, USA). The Lychnis RNA (29.7 μg) was sequenced on the Illumina HiSeq 2000 platform at LabGenomics (Seongnam, South Korea), generating 6 Gb of 100 bp of paired-end reads.

Transcriptome assembly and microsatellite identification

RNA sequence reads were assembled with Trinity v2.2.0 [48] on a 12-core 3.33-GHz Linux Workstation with 192 GB memory. These assembled contigs were used to detect microsatellite loci using the MIcroSAtellite identification tool (MISA) [49] with the following criteria: mono-nucleotide repeat motifs with at least 12 repeats, di-nucleotide repeat motifs with at least six repeats, tri- and tetra-nucleotide repeat motifs with at least five motifs, and penta- and hexa-nucleotide repeat motifs with at least four repeats. The criterion for compound microsatellites was that the interval between two repeat motifs was shorter than 100 nt. The assembled contigs were deposited in the Dryad Digital Repository (doi:

To investigate the distribution of microsatellites in the L. kiusiana leaf transcriptome, candidate coding regions within transcript sequences were identified using TransDecoder v3.0.1 ( The location of microsatellite was determined based on the predicted CDS region, 5′ UTR and 3′ UTR. Characteristics of motif type were determined and compared with each other among microsatellite loci located in the CDS region, 5′ UTR and 3′ UTR.

Functional annotation

To understand the possible function of microsatellites, all the transcripts harboring a microsatellite were searched against the GenBank nr protein database using BLASTx with an e-value cut-off of 10− 5 using Blast2GO v4.1.9 [50]. Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) were used.

SNP identification and estimation of synonymous and nonsynonymous sites

To identify putative SNPs in the transcripts containing SSRs, all reads were mapped onto the assembled transcripts using Bowtie v2.2.9 [51] with default parameters. The SAMtools “mpileup” utility and BCFtools [52] were used to identify SNPs. The SNPs with a read depth ≥ 10 and mapping quality ≥20 were retained. To detect transcriptomic SSRs under directional selection, nonsynonymous (dN) and synonymous (dS) rates of ORFs that contained at least one SNP were estimated by KaKs_Calculator v2.0 [53]. A dN/dS ratio greater than one suggests positive selection and less than one indicates purifying selection [54]. We estimated that ORFs with dN/dS > 1 and p < 0.05 (Fisher’s exact test) were under positive selection.

Microsatellite detection and genotyping

Previously extracted genomic DNA [17] for 51 individuals collected from the population of Minamioguni (MG, Kumamoto, Japan) was used. To explore amplification and polymorphism of the microsatellite loci identified in the L. kiusiana leaf transcriptome, the primers were designed based on the sequences flanking the microsatellite loci using Primer3 [55]. Apart from transcriptomic SSR markers under selection, 50 primer pairs were selected from the transcripts with a large number of repeat sizes and then used to amplify the microsatellite loci. Polymerase chain reaction (PCR) was performed in a 25-μl reaction volume containing 10 ng genomic DNA, 0.2 mM of each dNTP, 0.25 μM of each primer, 5 μl 10× h-Taq reaction buffer, and 0.25 units of h-Taq polymerase (Solgent Co., Daejeon, South Korea). PCR was performed under the following conditions: an initial denaturation at 95 °C for 15 min followed by 30 cycles each of 95 °C for 20 s, 58 °C or 60 °C for 40 s, and 72 °C for 30 s, and a final extension at 72 °C for 5 min. Electrophoresis of the amplified fragments was performed on 2% agarose gel to confirm the success of amplification and polymorphism. Forward or reverse primers selected for the population genetic analyses (Table S7) were labeled with 6-FAM, HEX, or TAMRA fluorescent dyes (Macrogen Inc.; Seoul, South Korea). For fragment analyses, PCR products amplified with fluorescent primers were genotyped on an ABI 3730xl DNA Analyzer (Applied Biosystems, California, USA) at Macrogen Inc. (Seoul, South Korea).

Microsatellite polymorphisms of L. kiusiana were calculated based on the average number of alleles per locus (NA), observed (HO) and expected (HE) heterozygosities and inbreeding coefficient (FIS) [56] in FSTAT 2.9.3 [57] and in GENEPOP v4.2 [58]. Departure from genotypic proportions expected under Hardy-Weinberg equilibrium (HWE) was examined by Fisher’s exact test using permutations [59] implemented in GENEPOP. The linkage disequilibrium between loci was also detected using Fisher’s exact test under the Markov chain algorithm implemented in GENEPOP. PIC was calculated using CERVUS 3.0 [60]. Critical significance levels were adjusted for multiple comparisons using the false discovery rate [61]. The statistical analysis was conducted using R v.3.3.3 [62].

Cross-species amplification

Three species of the genus Lychnis (L. cognate, L. fulgens, and L. wilfordii) were chosen to evaluate the transferability of these newly developed microsatellite markers to other related species. Total genomic DNA was isolated from herbarium specimens using the methods of Allen et al. [63]. PCR was performed as described above with the primer pairs. Electrophoresis of PCR products was performed on 2% agarose gel to check the success of amplification. The experiments were repeated three times with the same controls to confirm reproducibility and consistency.