2024 Volume 93 Issue 3 Pages 294-302
In Japan, wild evergreen azalea species with high ornamental value, such as Rhododendron ripense, grow naturally, and numerous cultivars have been developed based on these species. In this study, we utilized microsatellite markers to examine the genetic relationships among wild evergreen azalea species in Japan, particularly focusing on R. ripense, as well as assessing the genetic diversity of R. ripense. STRUCTURE analysis revealed that when K = 2, R. ripense appeared to be distinct from other species. However, when K = 4, the R. ripense population exhibited two separate clusters. Further analysis at K = 10 revealed genetic diversity within the R. ripense population, which was divided into five clusters reflecting their respective geographic distributions. Analysis of cultivars related to R. ripense based on the results of wild species suggested that most of the Ryūkyū-tsutsuji (R. × mucronatum) cultivars originated from the Yamakuni River in the northern Kyushu region or the San’in region of R. ripense.
In the genus Rhododendron L. (Ericaceae), which comprises over 1,000 species mainly in the tropical and arctic zones of the northern hemisphere, about 50 Rhododendron species are native to Japan (Kurashige, 2017). Evergreen azaleas in the subgenus Tsutsusi section Tsutsusi are indispensable genetic resources for cultivars used as ornamental shrubs or pot plants in temperate regions around the world. Several endemic Japanese wild species have high ornamental value, including small-flowered species such as R. kaempferi Planch., R. kiusianum Makino, R. indicum (L.) Sweet, R. eriocarpum (Hayata) Nakai, belonging to the subsection Tsutsusi, as well as large-flowered species such as R. scabrum G. Don, R. macrosepalum Maxim. and R. ripense Makino, which belong to the subsection Scabra. Cultivars have been bred based on these species since the Edo era (1603–1867). (Kobayashi, 2020; Kurashige and Kobayashi, 2008).
In particular, PCR-RFLP analysis of chloroplast DNA (cpDNA) suggested that R. ripense was an important wild species that contributed to the establishment of large-flowered cultivars etc. (Kobayashi et al., 2021). Furthermore, this species is highly adaptable to environmental conditions such as moisture and drought resistance, and the Ryūkyū-tsutsuji and Ōkirishima cultivar groups, which are presumed to be related to R. ripense, are used as street trees and public greening plants (Kobayashi et al., 2010a, b; Kunishige, 1976).
The distribution of R. ripense has been confirmed only from the Asahi River in Okayama prefecture, across the Seto Inland Sea to the Shikoku region and part of the Yamakuni River in the north Kyushu region. It is a semi-evergreen shrub with a height of 0.5 to 1.0 m. It grows naturally on riverside rocks in the upper and middle reaches of rivers (Kondo et al., 2009; Kurashige, 2017). The habitat of this species is in decline due to river improvement activities and human collection, resulting in its listing as endangered in various prefectures. Specifically, R. ripense is classified as critically endangered in Fukuoka prefecture, endangered in Oita prefecture, and vulnerable to extinction in Okayama prefecture (Fukuoka Prefecture Government, 2011; Oita Prefecture Government, 2011; Okayama Prefecture Government, 2020).
Studies evaluating the genetic diversity of R. ripense have been conducted using cpDNA analysis (Kobayashi et al., 2008) and microsatellite marker analysis (Kobayashi et al., 2017; Kondo et al., 2009). However, the genetic relationships between R. ripense and other wild species, as well as the genetic diversity among different provenances, remain unclear.
In this study, our objectives were to investigate the genetic relationships of R. ripense with other wild species and assess the genetic diversity among different R. ripense habitats. Additionally, we aimed to determine the provenances associated with the seven cultivars that are genetically and morphologically considered closely related to R. ripense. We conducted Bayesian clustering analysis using microsatellite markers on wild species and cultivar groups of evergreen azaleas, which are thought to have contributed to the establishment of cultivars, with a focus on R. ripense in each region, to examine the genetic relationships among wild species and the genetic involvement of wild species in cultivar groups.
In this study, nine species of evergreen azalea species (R. ripense, R. scabrum, R. macrosepalum, R. yedoense Maxim. ex Regel var. yedoense f. poukahanense (H. Lév.) Sugim. ex T. Yamaz., R. eriocarpum, R. indicum, R. simsii Planch., R. kaempferi, and R. kiusianum), predominantly native to Japan, were utilized, comprising a total of 432 individuals (Table 1; Fig. 1). These included samples used in Kobayashi et al. (2021). Seven cultivars related to R. ripense were collected from the Koishikawa Botanical Garden of the University of Tokyo and the Plant Breeding Laboratory, Faculty of Life and Environmental Sciences of Shimane University (Table 1).
Evergreen azalea wild species and cultivars used in this study.
Location of the sampled rivers inhabited by R. ripense.
Total genomic DNA was extracted from approximately 70 mg of −80°C freeze fresh leaves of each plant by a modified CTAB method (Kobayashi et al., 1998).
Microsatellite analysisA total of seven loci, RM2D2, AZA002, AZA003, AZA008, AZA010, AZA011 and N25 (Dendauw et al., 2001; Naito et al., 1998; Tan et al., 2009) developed from R. japonoheptamerum var. hondoense and R. simsii, were used as microsatellite primers (Table 2).
Primers used for microsatellite analysis.
Microsatellite primers were used with fluorescence (FAM, PET or VIC) on the 5' side of the forward primers and the tail on the 5' side of the reverse primers. PCR was performed in 10 μL reaction mixtures containing 5 ng of genomic DNA, 1 × reaction buffer (Contains MgCl2), 0.2 mM dNTPs, 0.25 U of Blend taq (Toyobo Co., Ltd., Osaka, Japan), and 0.2 μM of each primer. PCR amplification was conducted using a Gene atlas 482 (ASTEC Co., Ltd., Fukuoka, Japan) thermal cycler. PCR amplification reaction conditions for the five AZA primers were as follows, with some modifications to the method of Scariot et al. (2007): preheating at 94°C for 3 min; 35 or 38 cycles of denaturation at 94°C for 30 s, annealing at 55°C for 30 s (AZA003, 008) or 45 s (AZA002, 010, 011), and extension at 72°C for 30 s; and final extension at 72°C for 10 min (Gobara et al., 2017). The RM2D2 and N25 conditions were set up according to the method of Aizawa et al. (2016). The fragment length of PCR amplified products was determined by comparison with the GeneScan 500LIZ Size Standard (Applied Biosystems, CA, USA) using an ABI PRISM 3100xl Genetic Analyzer (Applied Biosystems) and Peak scanner ver. 3.0.3-PRC-build4 (Applied Biosystems).
Statistical analysesTo ascertain if each locus satisfied the prerequisites for population genetic analysis, we used FSTAT 2.9.3.2 (Goudet, 2002) to assess deviations from Hardy-Weinberg equilibrium (HWE) at individual loci and to examine linkage disequilibrium among loci within each population. Significance levels were adjusted using the Bonferroni correction for multiple tests. The mean number of alleles (NA), observed heterozygosity (HO), expected heterozygosity (HE), allelic richness (RA; El Mousadik and Petit, 1996), and fixation index (FIS) were calculated from the detected genotypes using GenAlEx6.51b2 (Peakall and Smouse, 2012) and FSTAT 2.9.3.2. A dendrogram was constructed to interpret the relationships among populations using a neighbor-joining (NJ) method based on the genetic distance, DA (Nei et al., 1983), using POPULATIONS 1.2.32 (Langella, 1999). To estimate the node significance of the dendrogram, we bootstrapped 1,000 replicates.
In order to estimate the origin of related R. ripense cultivars, we used STRUCTURE ver. 2.3.4 (hereafter, STRUCTURE analysis; Pritchard et al., 2000), referring to the analysis method for wild and cultivated Japanese flowering cherry (Kato et al., 2014). The STRUCTURE analysis was performed in two steps. First, in order to obtain data suitable for estimating the origin of cultivars related to R. ripense, we performed STRUCTURE analysis using only wild species individuals. The number of genetic clusters (K) was set from 1 to 22 under an admixture model with correlated alle frequencies. Each run involved 100,000 Markov chain Monte Carlo generations, after a burn-in period of 100,000 iterations. Fifty runs were performed for each K value. The optimal K value was determined log-likelihoods and ΔK (Evanno et al., 2005) using Structure Harvester ver. 0.6.94 (Earl and von Holdt, 2012). The average membership coefficients (q values) were calculated using CLUMPP 1.1.2 (Jakobsson and Rosenberg, 2007). CLUMPP implements the greedy algorithm, which involves switching labels and determining from each run whether a cluster corresponds to a specific label. The analysis results were visualized using Distruct1.1 (Rosenberg, 2004). The q values were derived from the top 10 runs with the highest Ln P(D) value among 50 replicates of the inferred K. Next, we examined the genetic relationship of wild species to R. ripense-related cultivars using the USEPOPINFO option in the STRUCTURE analysis. STRUCTURE analysis was performed using pre-clustering information for all individuals of the selected wild species (POPFLAG = 1), and the seven cultivars were of unknown origin (POPFLAG = 0). K was fixed at the optimal value determined in the previous step, and the genetic composition of POPFLAG = 0 was assessed based on the genotype information of POPFLAG = 1. Individual wild species used for USEPOPINFO were those with q values ≥ 80% in the most dominant clusters.
Microsatellite analysis was performed on the wild species (Table 1) using seven microsatellite loci. The results yielded a total of 296 alleles, with fragment sizes ranging from 127–325 bp and the number of alleles (NA) at each locus ranged from 23–76 (Table 2, S1, and S2). Testing for deviations from HWE was carried out for all combinations of nine populations and seven loci, and no deviations from HWE were detected.
The genetic diversity among the nine populations varied among wild species. The expected heterozygosity (HE) ranged from 0.664 in R. ripense to 0.906 for R. kaempferi, while the allelic richness (AR) ranged from 4.219 in R. ripense to 6.095 in R. kaempferi (Table 3).
Genetic diversity estimated from microsatellite loci in wild evergreen azalea species.
The NJ dendrogram summarizes the genetic relationships among wild species (Fig. 2). The dendrogram among wild species divided them into three large groups: An R. ripense group, large-flowered species (R. macrosepalum, R. scabrum, and R. yedoense), and small-flowered species (R. eriocarpum, R. indicum, R. simsii, R. kaempferi, and R. kiusianum). Small-flowered species that formed clades with high reliability included R. kaempferi (Kyushu region) and R. kiusianum, and R. eriocarpum, and R. simsii.
Dendrogram of wild species based on the neighbor-joining (NJ) method. The bootstrap values of over 50% based on 1,000 replicas are shown along branches.
In the STRUCTURE analysis results using 432 individuals from wild species, the optimal values of K were K = 2, 4, or 10 determined from ΔK and the log-likelihood (Fig. S1). Bar charts for the percentage of q-values are therefore presented for K = 2, 4, and 10 (Fig. 3). The proportion of values (q-values) derived from each individual is summarized for each wild species group (Table 4). At K = 2, the R. ripense population was distinctly separated from the other wild populations. At K = 4, the R. ripense population was divided into two distinct groups: one group consisting of the Yamakuni River and San’in region populations, and another group comprising the San’yo region and Shikoku region populations. On the other hand, the other wild populations were separated into two clusters: one cluster consisted of large-flowered wild species such as R. scabrum and R. macrosepalum, and the other cluster consisted of small-flowered wild species like R. kaempferi. Among others, a portion of the population assigned to the R. ripense cluster, specifically the San’in region and Yamakuni River populations, exhibited an admixture of clusters associated with R. ripense from the San’yo region and Shikoku region. At K = 10, the analysis demonstrated that the R. ripense population was further divided into five clusters: the Yamakuni River, San’in region, Takahashi River, Asahi River, and Shikoku region. Similarly, the other wild species were also grouped into five clusters: R. scabrum, R. macrosepalum, and three clusters representing small-flowered wild species.
Bar chart of 432 wild species individuals at K = 2, 4, and 10.
The q value of K = 2, 4, and 10 for each wild species estimated by STRUCTURE analysis.
For the STRUCTURE analysis attempting to estimate the genetic composition of R. ripense-related cultivars, we included 423, 375, and 271 individuals out of the total 432 wild species with q-values ≥ 80% at K = 2, 4, and 10, respectively. The Genetic composition of R. ripense related cultivars was estimated by referencing the genotypes of wild species, and a bar chart was generated to illustrate the results (Fig. 4). At K = 2, all cultivars were assigned to the R. ripense cluster with a percentage of over 85%. At K = 4, the White flower form of Kishi-tsutsuji and four Ryūkyū-tsutsuji (R. × mucronatum) cultivars were predominantly assigned to the R. ripense cluster (Yamakuni River, San’in region). On the other hand, the two remaining Kishi-tsutsuji cultivars were classified as R. ripense (San’yo region, Shikoku region). At K = 10, the results revealed that ‘Momoka’ and ‘Wakasagi’ were primarily associated with the Shikoku region within the R. ripense cluster. In the case of ‘Momoka’, genetic involvement with R. indicum was indicated. For ‘Wakasagi’, the genetic involvement was suggested with R. kaempferi. Among other cultivars, the white flower form of Kishi-tsutsuji, ‘Fujiman-yo’, and ‘Shiro-ryūkyū’ were predominantly associated with clusters from the Yamakuni River. On the other hand, ‘Murasaki-ryūkyū’ and ‘Shikinomai’ showed a dominance of clusters from the San’in region.
Bar chart for K = 2, 4, and 10 of cultivars based on wild species.
The allelic richness (AR) of R. ripense obtained from microsatellite markers was 4.219, which was lower than that of other wild species. The lower values observed for R. ripense compared to other wild species are likely attributable to its geographical isolation and distinctive distribution region (Kondo et al., 2009).
In STRUCTURE analysis results at K = 2, R. ripense and other wild species were clearly separated, suggesting that R. ripense underwent relatively recent speciation from another closely related species, R. macrosepalum (Kondo et al., 2009). Furthermore, the analysis indicated that R. ripense has independently adapted to river environments, distinct from R. indicum, which also naturally grows in river habitats (Kondo et al., 2009).
At K = 4, the analysis revealed a division into two clusters representing R. ripense and two clusters representing large-flowered wild species, specifically R. scabrum and R. macrosepalum, as well as small-flowered wild species such as R. kaempferi. According to the morphology-based classification proposed by Yamazaki (1996), within the section Tsutsusi, small-flowered wild species such as R. indicum and R. kaempferi are classified under the subsection Tsutsusi. In contrast, the large-flowered wild species R. ripense belongs to the Subsection Scabra, which also includes R. scabrum and R. macrosepalum. In this study, the genetic analysis revealed that R. indicum and R. kaempferi, which are classified under the subsection Tsutsusi according to Yamazaki’s (1996) classification, exhibited a close genetic relationship. However, in the subsection Scabra, R. ripense did not show genetic similarity to the other species in the same cluster. The dendrogram also revealed that R. ripense is genetically distant from R. scabrum and R. macrosepalum, reflecting the results of the structure analysis. Furthermore, in the results at K = 4 and 10, there were observations of population differentiation within the R. ripense population and the presence of cluster admixture within the R. ripense (San’in) cluster. Kondo et al. (2009) suggested that the formation of the distribution range and gene flow of R. ripense occurred independently in two ancient river systems that existed in the Chugoku and Shikoku regions prior to the formation of the Seto Inland Sea. Chloroplast DNA analysis of R. ripense native to the San’in and Shikoku region revealed differences in nucleotide sequences in the trnW–trnP region, suggesting regional variation in chloroplast DNA. Additionally, mixed sequences were observed in the Shikoku region (Kobayashi et al., 2008). The results of this study, along with previous reports, support the distinction of lineages of R. ripense in the San’in and Shikoku regions.
Based on the obtained genetic characteristics of R. ripense populations in different rivers, we evaluated R. ripense-related cultivars. Based on the K = 2 results, it became evident that the genetic composition of the cultivars used in this study was predominantly attributable to R. ripense. Kishi-tsutsuji ‘Wakasagi’, is an old R. ripense cultivar with thin petals and leaves. However, the results of the K = 2 analysis suggest that ‘Wakasagi’ contains genetic components from other wild species. Therefore, it is thought that ‘Wakasagi’ is an interspecific hybrid between R. ripense and other wild species. Kisi-tsutsuji ‘Momoka’ is marketed by nurseries as a cultivar with a trait of long-lasting flower corollas (misome-shō), and is derived from R. ripense from the Shikoku region. The results of this study suggest that the majority of the genetic components of ‘Momoka’ consisted of R. ripense from Shikoku, with the involvement of other wild species such as R. macrosepalum and R. indicum.
Ryūkyū-tsutsuji (R. × mucronatum) is related to R. macrosepalum and R. ripense, and is likely a hybrid of the two species (Inobe, 1971; Yamazaki and Yamazaki, 1972). ‘Fujiman-yo’ is an ancient cultivar recorded in the Edo period book “Kinshū-makura”. Yamazaki and Yamazaki (1972) reported that ‘Fujiman-yo’ exhibits a morphology similar to that of R. ripense. Chamberlain and Rae (1990) mentioned that R. mucronatum (Blume) G. Don var. mucronatum is a widely cultivated white form of R. mucronatum var. ripense (Syn. R. ripense), and it is possible that this white form occurs in the wild as the albino form of R. ripense. This white form of R. ripense is believed to correspond to the cultivar known as ‘Shiro-ryūkyū’. The results of this study suggest that the genetic contribution of R. ripense from the Yamakuni River or San’in region is implicated in the formation of ‘Shiro-ryūkyū’. ‘Murasaki-ryūkyū’ is thought to be a cultivar that has arisen through natural hybridization between R. ripense and R. macrosepalum within the Shikoku region (Inobe, 1971). The results of this study indicated that the genetic composition of ‘Murasaki-ryūkyū’ was largely R. ripense and it was suggested that R. ripense in the San’in region was also closely associated. The results of this study indicated that the majority of the genetic components of Ryūkyū-tsutsuji, including ‘Murasaki-ryūkyū’ and ‘Shiro-ryūkyū’ were derived from R. ripense. They also suggest a substantial genetic contribution from either the Yamakuni River or the San’in region, indicating the involvement of either of these R. ripense populations. In contrast, the genetic contribution from the San’yo and Shikoku regions was relatively low.
Here, we were able to evaluate the independence of R. ripense and the regional genetic diversity within R. ripense populations by using microsatellite analysis of seven loci. Furthermore, using these markers, we conducted an analysis of traditional cultivars, revealing that in the formation of the Ryūkyū-tsutsuji cultivar group, there was a significant genetic contribution from R. ripense populations in the Yamakuni River of northern Kyushu and the San’in region. The methodology of the STRUCTURE analysis used in this study will enable a more precise evaluation of the genetic contribution of wild species to cultivars such as Hirado-tsutsuji (R. × pulchrum), Ōkirishima (R. × pulchrum), and the Azalea cultivar group, which harbor cpDNA of the R. ripense type, as demonstrated in the study conducted by Kobayashi et al. (2021). In the future, the use of additional markers such as SNP markers derived from the R. ripense genome (Shirasawa et al., 2021) alongside microsatellite markers will allow for a more detailed elucidation of the contribution of R. ripense and other wild species to the development of cultivars.
We are grateful to Koishikawa Botanical Gardens, University of Tokyo for providing us with some of the azalea cultivars. The authors also thank the faculty of Life and Environmental Sciences in Shimane University for financial support in publishing this report.