Genetic analysis and population structure of wild and cultivated wishbone flower (Torenia fournieri Lind.) lines related to specific floral color

Introduction

The genus Torenia belongs to the class Magnoliopsida, order Scrophulariales and family Scrophulariaceae. There are more than 50 species native to tropical and subtropical Asia and Africa (Yamazaki, 1985; Hsieh & Yang, 2002). According to Yamazaki (1985), the significant biodiversity regions for this genus have been found in Southeast Asia (19 in Thailand, 20 in Cambodia, Laos, and Vietnam) and southern China (Hsieh & Yang, 2002). Wishbone flower or Torenia is the common name for several species in the genus including Torenia fournieri Lind., T. concolor Lindl., T. asiatica L. and T. hybrida (T. fournieri × T. concolor) (OGTR, 2008). T. fournieri, an annual from tropical Indochina, is one of the most important species in the genus for ornamental use in tropical and subtropical countries, especially as a bedding plant during summer in temperate regions. An erect type cultivar of this particular species with a violet floral color was initially commercialized. Then in 1988, PanAmerican Seed (West Chicago, IL, USA) released the Crown series with pink, white, and reddish-purple color lines. Selected from an interspecific cross between T. fournieri and T. concolor in 1995, a trailing type cultivar called the Summer Wave series was later released by Suntory Ltd. (Osaka, Japan). For more than 30 years, many hybrids have been released as patents or trademark series with a full range of floral colors from white with yellow throats to blue, cobalt, lavender, and violet, making Torenia a popular ornamental plant on the international market (Nagase, 1997; Miyazaki & Ohsumi, 1997; Tamura & Miyazaki, 1999; Miyazaki, 1999, 2001, 2002, 2003; Suzuki et al., 2002; Iwaki, 2005). As a consequence, a thorough knowledge of the genetic background of both wild and cultivated Torenia could aid future selective breeding of this important genus.

Early research employed T. fournieri as an ideal model plant for micromanipulation of artificial fertilization. Two major characteristics of the Torenia species included the exceptional protruding embryo sac (Higashiyama et al., 1997) and the synergid cells with polypeptides secreted as pollen tube attractants during fertilization (Okuda et al., 2009) that enable the detailed observation of the fertilization process. Consequently, the micromanipulation of artificial fertilization on Torenia as originally developed by Keijzer, Reinders & Leferink-ten (1988) resulted some interspecific hybrids, has since been adopted in a wide range of practical plant gene biotechnological experimentations (Higashiyama et al., 1998, 2001, 2006; Wu, Qin & Zhao, 2008; Wu et al., 2008; Nakamura et al., 2010; Nishihara et al., 2018; Sasaki & Ohtsubo, 2020). However, the resulted interspecific hybrids require genetic analysis.

In relation to new variants, Tandon & Bhutani (1965) first reported the morphological and cytological descriptions of the colchicine-induced T. fournieri tetraploid plant, Kikuchi et al. (2005) later undertook cytogenetic analysis with the formation of Torenia interspecific hybrids. Kikuchi et al. (2007a) conducted a thorough evaluation of pollen tube growth in cross combinations between T. fournieri and 14 related species and some interspecific hybrids were attained. Specifically on the T. fournieri and T. baillonii hybrid, a wide range of publications relating to cytological variations, genome size, interspecific genomic affinity, karyotypic variation, centromere separation and associations during mitosis or meiosis occurred (Kikuchi et al., 2006; Kikuchi et al., 2007b; Nuntha et al., 2016; Nuntha et al., 2017). Wei (2010) completed some experimentation on inheritance of the floral color, and analyzed the barriers to distant hybridization between closely related species in the genus Torenia, whilst Liang (2014) outlined a practical method on selective and mutational breeding to further achieve interspecific polyploid hybrids. The most recent interspecific Torenia hybrid was obtained through ovule culture (Laojunta et al., 2019). However, there has been no population genetic analysis on interspecific hybrids of Torenia.

A significant issue in the commercialization of Torenia interspecific hybrids is floral color, and specific colors have been released as individual cultivars. Some early chemo-genetic investigations of the floral color of T. fournieri revealed the inheritance pattern of anthocyanin concentrations in their flowers, and similarly, the inheritance pattern of the carotenoids in the petals of T. baillonii and T. fournieri have been described (Endo, 1962; Lang, 1970; Hess, 1971). Suzuki et al. (2000) reported that hybrid floral color in Torenia could be modified by suppressing the genes for anthocyanin biosynthesis whilst such floral color was later shown to be controlled, expressed and modified by the flavonoid 3′-hydroxylase and flavone synthase II genes (Ueyama et al., 2002). Through RNAi suppression of the anthocyanidin synthase gene, Nakamura et al. (2006) achieved pure white flowers with a higher frequency and better stability than through the antisense and sense suppression method. Furthermore, the floral pigmentation in Torenia species could be mutated by heavy ion beam irradiation (Miyazaki et al., 2006). Additionally, Torenia species exhibited a high degree of transformation by way of the Agrobacterium system (Aida & Shibata, 1995), making them popular for transgenic studies (Aida et al., 2000a, 2000b; Aida, 2008; Nishihara et al., 2013; Masahiro et al., 2014). These molecular biotechnological research results significantly enhanced our knowledge about floral color gene functioning in the Torenia genus. Consequently, these research findings provide useful background information for the future plant genetic improvement of Torenia floral color in cultivated and wild populations.

Regarding genetic activity and homeotic genes in particular, we know that they can regulate the development of anatomical structures in various organisms during evolution. Bradley et al. (1993) reported that some complementary floral homeotic phenotypes resulted from the opposite orientations of a transposon at the plena locus. In addition, the first double flower phenotype in Japanese morning glory (Ipomoea nil) resulted from gene duplication through insertion of an En/Spm-related transposable element into a floral homeotic gene (Nitasaka, 2003). Similarly, the chimeric TCP3 repressor in Arabidopsis was also found to be responsible for novel floral traits in T. fournieri and Chrysanthemum morifolium (Narumi et al., 2011). For floral gene expression in T. fournieri, the novel paracorollas trait has been generated when the floral homeotic genes were induced by the plant hormone forchlorfenuron (Niki et al., 2012) because the relative cytokinin biosynthesis system promoted a triggering expression of the cytokinin oxidase/dehydrogenase gene during such treatments (Niki et al., 2013a; Niki et al., 2013b). Such research findings enable us to conclude that some transposons in the homeotic genes can be the major genetic factors regulating certain phenotypic traits of the flowers. Therefore, use of transposon marker systems could be valuable in facilitating the population genetic analysis of Torenia in cultivated and wild populations so as to understand their evolutionary history as well as the selective pressure due to breeding for floral colors, in particular.

Specifically, for a practical transposon marker system, Kalendar et al. (2010) designed the iPBS (inter-Primer Binding Site) to deal with the “jumping genes” or transposable sequences of DNA that move from one location to another in the genome. This molecular marker technology utilized the universal presence of a tRNA complement as a reverse transcriptase primer binding site in LTR (Long Terminal Repeat) retrotransposons (Alzohairy et al., 2014). Different to the codominant marker systems such as microsatellite, iPBS is a dominant marker system that requires no previous knowledge of the genome and it is a more powerful DNA genotyping technique than the transposon-based markers established previously (Kalendar, Amenov & Daniyarov, 2019; Androsiuk et al., 2019). This marker system has been successfully applied to a wide range of plants such as ornamentals (Luo et al., 2015; Zhou et al., 2021), fruits (Baránek et al., 2012; Guo et al., 2014; Mehmood et al., 2016; Al-Najm et al., 2016), medicinal plants (Chen & Liu, 2014; Borna et al., 2017), and other different species (Singha, Nandhaa & Singha, 2017; Androsiuk et al., 2019; Androsiuk et al., 2020). Using such a retrotransposon-based iPBS technique, our objectives were to investigate the genetic pattern, population structure and genetic relationships among 136 Torenia accessions including 15 cultivated lines from three commercial series (Duchess, Kauai, Little Kiss) and two wild T. concolor populations sampled from two different sizes of area. Our overall aim was: (i) to estimate the genetic variation and population structure for T. fournieri cultivated lines which could be influenced by breeding selection; (ii) to explain the variation of the genetic diversity between the two wild populations could be differed from the sampling method we used; (iii) to provide insight into the effect on genetic variation in a population experiencing a bottleneck in which strong phenotypic selection overlaps; (iv) to provide a practical model for the genetic analysis of Torenia populations so as to better enable selective breeding in the future.

Materials and Methods

Material

The list of our 17 lines/populations included A (Duchess: Pink), B (Duchess: Burgundy), C (Duchess: Deep Blue), D (Kauai: Burgundy), E (Kauai: Rose), F (Kauai: Deep Blue), G (Kauai: Blue and White), H (Kauai: Magenta), I (Kauai: Lemon Drop), J (Kauai: White), K (Little Kiss: White), L (Little Kiss: Burgundy), M (Little Kiss: Blue and White), N (Little Kiss: Rose Picotee), P (Little Kiss: Blue), Q (wild: Lipu) and R (wild: Xichou). The origin of the plant materials used in this study are listed in File S1. We purchased a total of 15 T. hybrida seed lines from Shanghai Yuanyi Seedling Co. Ltd. These 15 colorful seed lines included three commercial series. Two wild T. concolor populations (Q: Lipu and R: Xichou) were collected from two different subtropical habitats in southern China, Lipu in Guangxi Zhuang Autonomous Region and Xichou in Yunnan Province (Fig. 1). To represent all the commercial lines that have been produced by selective breeding and currently released on the international market, approximately 200 seedlings for each line were germinated first, and eight plants were randomly selected to form a line. According to the availability of individual T. concolor plant in natural habitats, eight individual accessions of the population Q were collected from a small area with a minimum distance of 50 m, whilst population R was selected across a larger area with a minimum distance of 100 m. Commercial seeds were germinated and grown in a glasshouse under a controlled environment at the Flower Research Institute, Guangxi Academy of Agricultural Sciences in 2019. The floral colors for 17 populations were presented in Fig. 2. With a total of 136 accessions, 15 cultivated lines and two T. concolor wild populations were then used for iPBS analysis.

Scoring and analyzing the iPBS data

For the process of scoring, the DNA bands were carefully sized, visually examined and all faint bands were excluded. Only clear bands were included and same sized bands were treated as a single locus. The data record for each locus used ‘1’ for presence of a band and ‘0’ for absence so as to form a binary matrix. The polymorphic informative content (PIC) value for each primer was calculated by PIC = 1 − [f² + (1 − f)²], where f is the frequency of the dominant marker in the data set. Accountable bands consistently appeared in all the repeated experiments.

Accessions with different names that were completely matched at all the loci were considered as duplicates. Duplicates in the data set were checked by multi-locus matching. All analyzes were performed with the GenAIEx 6.5 program (Peakall & Smouse, 2012) (Fig. 3). Pair-wise genetic distance was first calculated using the DISTANCE procedure. Later analysis included principal coordinates analysis (PCoA), and data were collected also as previously described in Peakall & Smouse (2012). The individual based pairwise distance matrices for populations were also used for hierarchical AMOVA analysis, and pairwise PhiPT values from the Phi-statistics were used to measure population differentiation due to genetic structure estimated from the genetic polymorphism data. Each time, the data were subjected to 1,000 permutations. As an indication of the relative contribution of population differentiation, the greater the pairwise PhiPT values mean the greater the differences between populations. The total genetic variation was partitioned across three levels: within populations (Phi-PT), among populations within taxa (Phi-PR), and among taxa (Phi-RT).

For dendrogram construction, data were collected and used the MEGA7.0.14 software (Kumar, Stecher & Tamura, 2016) for analysis as previously described in Zhou et al. (2021). There was a total of 148 positions for our 136 taxa in the final dataset. Using the UPGMA method (Sneath & Sokal, 1973), the optimal tree with the sum of branch length = 5.0639 is shown. The final tree is drawn to scale, with branch lengths in the same units as those of the evolutionary distances used to infer the phylogenetic dendrogram. The evolutionary distances in the optimal tree were computed and are in the units of the number of base substitutions per site (Tamura, Nei & Kumar, 2004). A final dendrogram revealing the evolutionary relationship of 136 Torenia taxa in 17 lines/populations is listed in Fig. 4.

The genotypic data were also analyzed using a Bayesian model-based clustering method implemented through STRUCTURE V2.3.4 (Pritchard, Stephens & Donnelly, 2000) for population structure to classify 17 Torenia lines/populations into different groups. Data were collected as previously described in Zhou et al. (2021). In Fig. 5, the optimal number of clusters or the most likely K value (K = 5) was based on Evanno, Regnaut & Goudet (2005). Population structures were constructed using a Bayesian model-based clustering computation from STRUCTURE HARVESTER (Earl & Vonholdt, 2012), and a final population structure was optimized in Fig. 6.

Results

Discussion

A collection of 136 accessions of 15 cultivated lines and two wild T. concolor populations with specific floral color were investigated using seven retrotransposon-based iPBS primers, and our results are presented as a PCoA plot, a phylogenetic tree, and an optimized population genetic structure bar plot. All the 148 loci generated from seven selected primers were sufficiently powerful to discriminate genetically our individual accessions into the origin of the line/population, whilst indicating an obvious pattern of genetic relationships within and among the lines/populations. Results from these analyses, individually and combined, indicate that some morphological traits such as the floral colors could be used as an associated genotypic identification, thereby potentially guiding breeding practices that target such traits. This research also showed that the comparison of cultivated T. fournieri lines and T. concolor wild populations are informative, and these two different species are being compared when analyzing the results.

We know that population structure analyses using dominant markers such as the iPBS system can be weakened if the heterozygosity remains undetected or underestimated. To prevent such an issue in our analysis, we adopted Phi-statistics which are a modified version of the conventional Wright’s F-statistics. Taking into consideration the genealogical information from the molecular data, the pairwise population PhiPT values were applied to measure the reduction of the expected level of heterozygosity in the population compared to a Hardy-Weinberg expectation that is usually used. Specifically, the Phi-statistics analysis we used showed the correlation of some genes or loci within the subdivided lines/populations. In particular, the genetic pattern and structure of the wild population Q (selected within a relatively small area) was similar to the 15 cultivated lines, but was significantly different from the other wild population R (selected across a much larger area). This most likely indicates that these differences in genetic diversity were resulted from bottleneck/drift effects when bringing a small number of individuals into cultivation during breeding practices. Thus, this Phi-statistics method could be used further to compare the genotype data of the parent lines against the hybrid lines so as to estimate such bottleneck/drift effects mentioned above. On the other hand, it needs to be noted that when considering wild populations within their natural habitats, sampling methods are one of the major factors affecting the conclusions being drawn about the population genetic structure.

Since blue is the common wild type of floral color for T. fournieri and T. conclor, and our two wild T. concolor populations had blue flowers, all the cultivated lines were bred for a specific floral color. In our study, clustering information among cultivated lines and wild populations via our PCoA analysis showed interesting patterns, and much of the variance in genotype among lines is due to differences between certain sets of lines which are most distant from one another. For instance, among cultivated lines I, K, M, N, P and the two wild populations Q and R, there were similar reduction in diversity/heterozygosity from a bottleneck event, and each line/population will retain a similar (but different from each other) content of the wild genetic diversity. As Mehmood et al. (2014) has shown, it is highly probable that the relevant floral color of these lines/populations is associated with a negatively correlated genetic relationship with the wild genetic content mentioned above. Therefore, as the highest loadings in the first two components, strong positive correlations presented in lines A, C, E, M, N, and these cultivated lines have less content of the wild genetic diversity. Consequently, these results from our population genetic analyses could guide future Torenia breeding.

Being useful for reconstructing the evolutionary history of lines/populations from population genetic analysis, the population genetic distance can be interpreted as a measure of genetic divergence between species or between lines/populations within a species, whether the distance evaluate the time from common ancestor or the degree of differentiation. Generally, species or wild populations of a species with many similar alleles exhibit small genetic distances due to having had a recent common ancestor. Results from our study indicated that the wild population Q is genetically more closely related to lines M, N, P and they are likely therefore to have had an overall genetic similarity. For the purpose of expanding heterozygosity, these closely related lines/populations should not be used as parent candidates for any inter-line/population crosses, especially those targeting introgressive traits from wild populations. On the contrary, it would be better to cross these closely related lines/populations with other lines/populations containing more genetic differentiation such as A–F.

As previously indicated, this study revealed an extensive genetic diversity and structural pattern within and among the 17 Torenia lines/populations studied through the 136 accessions. A large number of specific bands deserve special attention as they were private alleles specific to the line(s)/population(s). The genetic information from these specific alleles indicated significant genetic relationships that separated certain lines/populations, as presented in the above PCoA analysis, phylogenetic dendrogram, and the final population structural bar plot. Furthermore, such population specific polymorphic bands can be excised, then purified from the gel and sequenced to provide more thorough genomic information relevant to the genetic analysis of the population. These specific molecular markers could also be associated with particular phenotypic traits of interest, such as floral color.

Regarding the heterozygosity for 17 Torenia populations tested, 15 commercial lines and the wild population Q had low values of Na (number of different alleles), Ne (the number of effective alleles), I (Shannon’s information index), He (expected heterozygosity), and uHe (unbiased expected heterozygosity). Line J has the lowest value of records whilst wild population R had the highest values. Since wild population Q was sampled from a smaller area, and population R was sampled across a much larger area, the different results of these two wild populations could be due to our sampling practices, as suggested above. Furthermore, the mean value of I (0.0758), He (0.0513) and uHe (0.0547) of all the 17 Torenia populations tested are considerably less than that reported previously in other genera of plants by Gailite & Rungis (2012), Guo et al. (2014), Baránek et al. (2012), Mehmood et al. (2016), Al-Najm et al. (2016) and Borna et al. (2017). Our lower values indicate significant low genetic diversity/heterozygosity in the 15 Torenia series that have been released, and in the wild population Q. This suggests either some kind of bottleneck/drift effect by selection during breeding practices.

Our results also support the view that the retrotransposon-based iPBS system we used could also be used to estimate genetic diversity and structure of other plant species. This would enable population genetic analysis on other biological phenomena such as adaptation, speciation, and population structure. Whilst Torenia species have been used previously for frontier research on fertilization and gene engineering of ornamental characteristics within horticultural ecosystems, our research conclusions could complement and expand its use as a model plant for new interests in plant population genetic analysis. Therefore, the results from this study could have two main flow-on effects. The first could be information applicable to selective breeding for commercial purposes. A second application could be with endangered species because the population genetic analysis would be able to index the strength of the population of the species in its natural habitat.

Conclusions

The iPBS markers showed interesting patterns of genetic variation among wild T. concolor populations and T. fournieri cultivated lines with different floral color. For cultivated lines and wild populations, there were similar reduction in diversity/heterozygosity from a bottleneck event, and each line/population retained a similar (but different from each other) content of the wild genetic diversity. Variances in genetic diversity/heterozygosity for the two wild populations could be explained by differences in sampling method. Environmental factors could induce transposon activation and generate genetic variability which enabled the acceleration of the evolutionary process of wild Torenia species, but selective breeding targeting a specific floral color decreased genetic diversity. As a consequence, this study also provide insight into the effect on genetic variation in a line/population experiencing a bottleneck in which strong phenotypic selection overlaps, but wild Torenia populations sampled from broad geographic region represent stronger species strength with outstanding genetic diversity. Using of the iPBS marker system not only revealed genetic variation and population structure for wild and cultivated Torenia plants, but this result could guide future selective breeding especially for commercial purposes.

Supplemental Information