INTRODUCTION
The genus Rhododendron L. (Ericaceae) comprises more than 1,200 species, making it one of the largest genera, and is widely distributed throughout the temperate and subtropical regions of the Northern Hemisphere (Bailey and Bailey, 1976; Park and Song, 2010). Within East Asia, 37 and 51 species have been reported in China and Japan, respectively, while 23 species are native to Korea (Hwang, 1998; Park and Song, 2010; Yang et al., 2015). Most species in the genus are deciduous shrubs, thriving across a broad elevational gradient from coastal lowlands to mountainous regions at altitudes of up to 2,000 m, and are thus widely distributed throughout the Korean Peninsula (Hwang, 1998). Plants typically reach a height of 2–3 m, with alternate leaves that bear slight scales on the surface (Choi et al., 2018). Flowering occurs from late March to early June, producing magenta to light pink flowers (Hwang, 1998; Choi et al., 2018). Morphological characteristics are strongly influenced by environmental factors, varying with bioclimatic conditions and the geographic location (Koksheeva et al., 2015). Considerable variation exists in traits such as the leaf morphology, flower structure, and growth habit, and substantial differences may occur even within the same species (Gibbs et al., 2011). Due to their diverse morphological traits and natural hybridization, Rhododendron species have for centuries been widely used for ornamental and horticultural purposes (Hwang, 1998; Blazich and Rowe, 2003; Gibbs et al., 2011). Today, thousands of cultivars are cultivated worldwide, and the genus continues to serve as an important subject of ecological and taxonomic research (Hwang, 1998; Wang et al., 2007; Park and Song, 2010; Gibbs et al., 2011).
In Korea, several species of the genus Rhododendron occur naturally, representing an important component of East Asian Rhododendron diversity. The deciduous species, including (Korean Rhododendron), R. mucronulatum Turcz, R. schlippenbachii Maxim.(royal azalea), and their variants R. mucronulatum var. ciliatum Nakai and R. yedoense f. poukhanense (H. Lév.) Sugim. ex T. Yamaz., are widely distributed across mountainous regions of the Korean Peninsula and possess considerable ornamental and ecological value. In contrast, R. micranthum Turcz, R. redowskianum Maxim., and R. lapponicum subsp. parvifolium (Adams) T. Yamaz. are northern or alpine species with restricted distributions, regarded as sensitive indicators of climate change. Furthermore, R. brachycarpum D. Don ex G. Don, R. aureum Georgi, and R. sohayakiense var. koreanum Y. Watan. & T. Yukawa occur in subalpine coniferous forests or rocky highlands, with R. sohayakiense var. koreanum recognized as an endemic taxon confined to Mt. Halla on Jeju Island. Collectively, these ten taxa exhibit distinctive ecological roles and distributional patterns, making them significant subjects for studies of the diversity, evolution, and conservation of East Asian Rhododendron.
Taxonomic studies of the genus Rhododendron have included palynological research (Park and Song, 2010), molecular phylogenetic analyses (Goetsch et al., 2005), and morphological investigations (Wang et al., 2007). Within studies focusing on the external morphology, seed morphological research is ongoing (Wang et al., 2014; Koksheeva et al., 2015; Namgay and Sridith, 2021). Despite this progress, previous studies have only addressed seed morphological characteristics to a limited extent, and research focusing on the seed surface structure and micromorphological traits remains scarce. In particular, studies of Rhododendron species distributed in East Asia are notably lacking in the literature. Seeds play a crucial role in plant reproduction and propagation, and seed morphology analyses are taxonomically significant, providing valuable information (Barthlott and Ehler, 1977). In particular, the structure of the exotesta has been shown to serve as an important criterion for species delimitation, as it represents a stable and conserved trait that is minimally influenced by environmental factors (Barthlott, 1981; Wang et al., 2007). Despite this significance, seed structure has not been sufficiently considered in phylogenetic studies (Davitashvili and Karrer, 2010; Ghimire et al., 2015, 2017, 2020; Rashid et al., 2021; Mazur et al., 2021).
Therefore, this study investigated the seed external morphological characteristics of ten species of Rhododendron distributed in East Asia. Seed size and weight were measured, overall seed morphology was examined using a 3D multifocus digital microscope, and the seed surface and micromorphological features were observed by means of scanning electron microscopy (SEM). Based on morphological characteristics and quantitative traits, statistical analyses were conducted to evaluate whether the analyzed traits contribute to species identification. Furthermore, the study aimed to clarify the seed morphological characteristics of Rhododendron, thereby assessing the taxonomic utility and limitations of seed morphological traits in species delimitation.
MATERIALS AND METHODS
Specimens
To compare the morphological traits of seeds, we examined 62 accessions representing ten species of Rhododendron. A total of 1,890 seeds were analyzed, with ten to thirty seeds used per accession. All seeds used in this study had undergone shade drying and a cleaning process to remove impurities, and all were stored in the Seed Bank of the Korea National Arboretum. A detailed list of specimens and collection information for each species is provided in Appendix 1.
Digital photography
The morphological characteristics of the seeds, in this case their color, size, and shape, were examined using a stereo microscope (MZ16 FA, Leica Microsystems, Wetzlar, Germany) and a 3D microscope (KH-8700, Hirox, Tokyo, Japan). Digital images of whole seeds were taken with a Leica DFC420C multifocal camera attached to a Leica MZ16 FA microscope (Leica Microsystems) and Hirox 3D microscope (Hirox). To observe the seed surface structure and micromorphological features, SEM was employed. Selected seeds were mounted on aluminum stubs using double-sided adhesive tape and sputter-coated with gold. Gold coating was performed using an ion coater (G-20DA, ULVAC, Chigasaki, Japan) operated at 4 mA for 200 s. The coated samples were examined with a COXEM CX-100S SEM device (COXEM, Daejeon, Korea). Overall seed morphology was observed at magnifications of 60–210×, while seed surfaces were examined at 1,000–1,500×. Seed weight was determined as 1,000-seed weight and an X-ray analysis served as a non-destructive method to assess the seed fill rate, enabling an evaluation of the correlation between seed weight and seed quality.
Presentation and description of the seed features
Approximately 1,890 seeds were used for morphometric measurements. Measurements of the seed length and width were taken with Leica LAS V3.8 software. The length-to-width ratio was subsequently calculated. All measurements are presented as the means ± standard deviations (SDs) and are summarized in Table 1.
Statistical analysis
All statistical analyses were performed using R Studio version 4.4.1 (Posit Software, Boston, MA, USA) and SPSS version 30.0 (IBM Corp., Armonk, NY, USA). In SPSS, the mean and SD of the seed morphological traits (length, width, length/width ratio, and 1,000-seed weight) were calculated. Welch’s one-way analysis of variance was conducted to test for interspecific differences, followed by the Games-Howell post-hoc test to identify significant pairwise differences among the species. Box plots were generated to visualize the distribution and variability of the seed traits across the species. In addition, principal component analysis (PCA) was conducted based on seed traits to evaluate the overall patterns of variation and species clustering.
RESULTS
The seed morphological traits of 62 populations representing the ten species of Rhododendron were examined and described. The external morphology was documented using stereomicroscopy and SEM, with representative images presented in Figs. 1–4. Seed shape, size, color, weight, and the presence are summarized in Table 1. By means of Welch’s one-way ANOVA and the Games-Howell post hoc test, the results of a statistical analysis of the mean differences in the sizes and weights of the ten species are shown in Appendices 2–4. Comparisons of seed surface and microscopic traits are summarized in Table 2.
Gross seed morphology
Most Rhododendron seeds examined in this study exhibited brownish hues, which could be categorized into four distinct color groups: red-brown, brown, orange-brown, and gray. Four species, R. micranthum, R. aureum, R. brachycarpum, and R. mucronulatum, displayed red-brown seeds. Three taxa, R. yedoense f. poukhanense, R. schlippenbachii, and R. mucronulatum var. ciliatum, produced brown seeds, while two species, R. lapponicum subsp. parvifolium and R. sohayakiense var. koreanum, had seeds showing orange-brown coloration. Finally, R. redowskianum was the only species with gray seeds (Fig. 1, Table 1). With regard to seed shape, all examined species predominantly exhibited an elliptical form. In addition, five taxa (R. mucronulatum, R. schlippenbachii, R. mucronulatum var. ciliatum, R. lapponicum subsp. parvifolium, and R. sohayakiense var. koreanum) also displayed obovate seeds. Rhododendron aureum seeds were ellipsoidal to oblong, R. yedoense f. poukhanense exhibited oblong seeds, and R. redowskianum presented cylindrical shaped seeds (Fig. 1, Table 1). Furthermore, basal and apical wings were observed at one end of the seeds in R. aureum and R. brachycarpum, whereas R. micranthum and R. mucronulatum possessed apical wings, which distinguish them morphologically from the other taxa (Fig. 2, Table 1).
Seed size
The seed morphological traits of the ten taxa of Rhododendron were analyzed and compared. Seed length, width, the length-to-width ratio, and the 1,000-seed weight differed significantly among the species (Fig. 5, Table 1, Appendices 2–4). Across all taxa, the mean seed length was 1.82 ± 0.48 mm, the mean width was 0.65 ± 0.17 mm, the mean length-to-width ratio was 2.89 ± 0.87, and the mean 1,000-seed weight was 0.1558 g. Rhododendron brachycarpum exhibited the longest seeds (2.44 ± 0.43 mm), showing a distinctly longer seed form compared to the other species (Table 1). This was followed by R. micranthum (2.23 ± 0.33 mm), R. schlippenbachii (1.97 ± 0.30 mm), and R. redowskianum (1.91 ± 0.24 mm), which also presented relatively high values (Table 1, Appendix 2). The shortest seeds were those of R. sohayakiense var. koreanum (1.34 ± 0.15 mm), followed by R. lapponicum subsp. parvifolium (1.43 ± 0.16 mm) (Table 1, Appendix 2). Regarding seed width, R. schlippenbachii had the widest seeds (0.82 ± 0.14 mm), clearly distinguished from the other species, followed by R. brachycarpum (0.72 ± 0.21 mm) and R. redowskianum (0.67 ± 0.08 mm) (Table 1, Appendix 3). The narrowest seeds were measured for R. lapponicum subsp. parvifolium (0.50 ± 0.07 mm), showing a significantly smaller value than the other species (Table 1, Appendix 3). The length-to-width ratio was highest for R. micranthum (3.92 ± 0.78) and R. brachycarpum (3.55 ± 0.90), whereas R. redowskianum exhibited the lowest ratio (2.18 ± 0.31) (Table 1, Appendix 4). Seed size variability was greatest in R. brachycarpum, showing the highest SDs in the length, width, and length-to-width ratio (Table 1). The heaviest seeds were observed in R. schlippenbachii (0.3238 g), and the lightest seeds were found in R. lapponicum subsp. parvifolium (0.0577 g) (Table 1). PCA based on seed size and weight showed that the first axis explained 41.50% of the variation and the second axis 22.21%, with a cumulative explanation rate of 63.71%. Seed size and weight varied significantly among the species, and PCA further confirmed these differences by forming three loose clusters along the first two axes. Group 1 comprised species with relatively larger seeds, including R. micranthum and R. brachycarpum, which exhibited elongated seeds with relatively narrow widths compared to their lengths. Group 2 consisted of species with distinctly smaller seeds, in this case R. mucronulatum, R. mucronulatum var. ciliatum, and R. sohayakiense var. koreanum, which were positioned at the opposite end of the ordination space. Group 3 was represented by R. schlippenbachii, which formed a distinct cluster due to its widest and heaviest seeds (Fig. 6). The remaining species were either positioned between groups or did not belong to a single cluster (Fig. 6).
Seed surface
The primary surface sculpture of Rhododendron species is predominantly striate, varying from irregular striate (R. micranthum, R. schlippenbachii), wavy striate (R. brachycarpum, R. yedoense f. poukhanense), and linear striate (R. aureum, R. mucronulatum, R. lapponicum subsp. parvifolium, R. sohayakiense var. koreanum) to reticulate (R. redowskianum, R. mucronulatum var. ciliatum) (Fig. 3, Table 2). Across all species, the surface cells are highly elongated with well-defined, raised boundaries. The periclinal wall is mostly flat and smooth (R. aureum, R. brachycarpum, R. yedoense f. poukhanense, R. mucronulatum, R. lapponicum subsp. parvifolium) (Fig. 4), occasionally smooth and undulating (R. schlippenbachii, R. sohayakiense var. koreanum), convex and granular (R. redowskianum, R. mucronulatum var. ciliatum), or slightly concave and wrinkled (R. micranthum) (Fig. 4, Table 2). The anticlinal walls are slightly to highly raised, displaying straight or nearly straight outlines in R. micranthum, R. aureum, R. brachycarpum, and R. mucronulatum; irregular or weakly sinuous outlines in R. yedoense f. poukhanense, R. schlippenbachii, R. lapponicum subsp. parvifolium, and R. sohayakiense var. koreanum; and strongly sinuous or undulate patterns in R. redowskianum and R. mucronulatum var. ciliatum (Fig. 4, Table 2). Species in the second group show moderate irregularities, resulting in asymmetric, less uniform cell patterns, whereas those in the third group exhibit deeply interlocked and highly undulating cell boundaries.
DISCUSSION
This study examined ten taxa of the genus Rhododendron, providing details about the seed morphology in each case through digital imaging. Our findings demonstrated that species within the genus can be distinguished based on a combination of seed color, size, weight, shape, and seed coat micro-morphological characteristics.
Seed color in Rhododendron was predominantly brown, with variations that included brown, red-brown, and orange-brown (Fig. 1, Table 1). In addition, certain Rhododendron species exhibit yellowish-brown or even yellow seed coloration. Species with yellowish-brown seeds include R. simsii, R. championiae ‘Fenhe,’ and R. simsii ‘Shenqiuhong,’ whereas R. molle is characterized by yellow seeds (Zhou et al., 2013). Whereas R. redowskianum exhibited unique gray coloration, the remaining species could be classified into four distinct color types (Fig. 1). The predominance of brown shades suggests that seed color is a conserved trait at the genus level (Wang et al., 2014). These differences in pigmentation are likely attributable to variations in pigment expression, also associated with adaptive ecological strategies (Nystrand and Granström, 1997; Zhao et al., 2022). At the same time, seed pigmentation can be environmentally modulated. In particular, high light or UV exposure levels may activate pigment-related metabolism, including anthocyanin accumulation, leading to relatively dark seed coloration (Kovinich et al., 2015). Drier or cooler conditions may likewise promote darker tones through the increased deposition of phenolic compounds in the seed coat (Kovinich et al., 2015). Consistent with this interpretation, dark-colored seeds have been reported to provide greater UV protection and confer crypsis in soil, thereby reducing the risk of predation (Porter, 2013; Jackson et al., 2021). In contrast, lighter-colored seeds tend to possess thinner seed coats and are more responsive to light fluctuations, facilitating rapid germination (Bhatt et al., 2016; Zhang et al., 2023). This tendency may be associated with lower levels of the accumulation of proanthocyanidins or differences in their composition and oxidation pathways, which can result in reduced seed coat thickness and less extensive structural or surface deposition (Qu et al., 2013). Consistently, lighter-colored seeds have been reported in other plant groups to exhibit a thinner palisade layer and fewer surface deposits, often accompanied by lower dormancy and a faster germination response (Zhang et al., 2023; Wen et al., 2024). Thus, seed color in Rhododendron should not merely be regarded as a morphological feature but rather as an ecologically functional trait linked to both dispersal and germination strategies.
Seed shape and appendages further contributed to interspecific distinctions. Most Rhododendron seeds lacked prominent wings and were typically elliptical, with occasional obovate or oblong forms (Fig. 2, Table 1). The prevalence of elliptical and oblong seeds across diverse elevations supports the interpretation that this represents a conserved form of seed morphology within the genus (Wang et al., 2014). Among the Rhododendron species not occurring in Korea, R. lanatum, R. cinnabarinum, R. setosum, and R. anthopogon typically bear small, elongate seeds characterized by an elliptic outline that gradually tapers toward both ends (Namgay and Sridith, 2021). In contrast, there are Rhododendron species that possess well-developed basal and apical wings, with the wings being separated from either the basal or apical region of the seed. Among species distributed in Korea, R. aureum and R. brachycarpum exhibit this morphology (Fig. 2). In these species, the apical wing is located at the proximal end of the seed, whereas the basal wing occurs at the distal end, and this arrangement resembles the morphological pattern observed in R. ponticum, R. nipponicum, R. arboretum, R. Makinoi, and R. wardii (Collinson and Crane, 1978; Hwang, 1998; Shalabi et al., 2020; Namgay and Sridith, 2021). In addition to possessing well-developed basal and apical wings, some species also develop distinct lateral wings. Examples include R. falconeri, R. dalhousiae var. rhabdotum, R. maddenii and R. molle (Zhou et al., 2013; Namgay and Sridith, 2021). These morphological features likely result from long-term selective pressures, with basal and apical wings playing a functional role in enhancing wind dispersal (Goetsch et al., 2011; Namgay and Sridith, 2021). By increasing the seed surface area and reducing the terminal velocity, these wing-like appendages facilitate prolonged suspension in the air, thereby enabling greater horizontal displacement from the maternal plant by prevailing air currents (Howe and Smallwood 1982; Hwang, 1998).
Seed size and weight varied significantly among the species, and PCA further confirmed these differences. R. micranthum and R. brachycarpum were placed in the same group due to their similarity in seed size and weight, positioning them among the relatively larger and more elongated seed types within the genus (Fig. 6, Table 1). In contrast, R. schlippenbachii formed a clearly separated cluster, distinguished by its broadest and heaviest seeds among the taxa analyzed (Fig. 6, Table 1). Although large and heavy seeds are generally associated with higher seedling survival under stressful conditions such as shade, drought, or mechanical disturbance owing to their greater nutrient reserves (Leishman and Westoby, 1994; Moles and Westoby, 2004), this interpretation is difficult to apply to the genus Rhododendron, as the results of this study are based on relative variation observed within the genus rather than on absolute differences in the seed size or weight. At the opposite end of the PCA distribution, R. lapponicum subsp. parvifolium and R. sohayakiense var. koreanum possessed markedly smaller seeds. Across all taxa, mean seed length and width were approximately 1.82 mm × 0.65 mm, with a 1,000-seed weight of 0.1558 g, indicating that Rhododendron seeds are generally very small and lightweight. This pattern is consistent with reports based on other Rhododendron species, such as R. kaempferi, R. kiusianum, and R. luteum, whose seed length typically ranges from 1.0 to 2.5 mm and width from 0.2 to 1.0 mm, while their 1,000-seed weight rarely exceeds 0.5 g (Hwang, 1998; Zhou et al., 2013; Shalabi et al., 2020). Particularly, R. sohayakiense var. koreanum exhibited exceptionally small and light seeds, a trait likely associated with its relatively small seedling size and growth in high-altitude regions (Hwang, 1998; Jakobsson and Eriksson, 2000; Wang et al., 2014). Nevertheless, the consistently small and lightweight seeds observed across the genus Rhododendron appear to represent natural variation within the group rather than a major functional divergence. In high-altitude environments, these seed traits may be shaped by resource constraints associated with short growing seasons (Baker, 1972; Wang et al., 2014; Namgay and Sridith, 2021), and they are advantageous for producing numerous propagules that can be effectively dispersed by wind, water, or animals (Rees, 1996; Moles et al., 2005). Moreover, small and lightweight seeds are generally more resistant to desiccation, a trait that may further minimize damage under the strong winds and highly exposed ridge environments characteristic of alpine habitats (Daws et al., 2006). Therefore, although relative differences exist among species, the shared ecological strategy of producing small, lightweight seeds can be regarded as a key adaptive feature of the genus (Hwang, 1998; Zhou et al., 2013; Shalabi et al., 2020).
From a phylogenetic perspective, the seed characteristics documented in this study, including variations in the size, weight, and overall morphology, do not necessarily correspond to the major clades identified in recent plastome-based and transcriptome-based nuclear phylogenies of Rhododendron (Mo et al., 2022; Xia et al., 2022). In the PCA analysis, R. micranthum and R. brachycarpum, which have relatively large and elongated seeds, were found to belong to different higher-level lineages in the molecular phylogenies (Xia et al., 2022). Similar patterns were observed in other taxa examined in this study, showing that similarity in seed traits does not necessarily reflect close phylogenetic relationship within the genus. These mismatches between seed-trait clusters and phylogenetic groupings are consistent with the findings of broader comparative work on Rhododendron showing that seed length, width, weight, and wing length vary significantly among species and subsections but show little differentiation among sections or subgenera (Wang et al., 2014).
The seed coat micro-morphology provides valuable diagnostic traits for species delimitation. The seed surface morphology in the Ericaceae case exhibits the characteristic striated antheridial ridges typical of angiosperms, forming patterns from striped to reticulated (Barthlott, 1981; Szkudlarz, 2001). Reticulated seeds, in particular, show differences in cell wall thickening and cell expansion among taxa, allowing for differentiation through other characteristics, such as the seed coat cell shape, the number of cell layers within the seed coat, and the radial curvature of the epidermal cells (Szkudlarz, 2001). Most taxa exhibit striate anticlinal ridges, a feature characteristic of angiosperms, with distinct interspecific variations including irregular striate, wavy-striate, linear-striate, and reticulate–striate patterns (Barthlott, 1981). For example, R. micranthum showed dense, continuous ridges, while R. aureum had smoother, linear ridges, despite both sharing a parallel ridge pattern. Rhododendron yedoense f. poukhanense exhibited irregularly branched ridges, and R. redowskianum was characterized by thickened, irregular ridges with prominent anticlinal walls (Figs. 3, 4, Table 2). In R. schlippenbachii, the deeply indented and thick anticlinal walls resulted in conspicuous surface irregularities. Notably, intraspecific variation was also observed, such as between R. mucronulatum and its variety R. mucronulatum var. ciliatum (Figs. 3, 4, Table 2). The former exhibited curved ridges with partial reticulation and a low to medium ridge density, whereas the latter had denser, wavy ridges with a more pronounced reticulate pattern and higher ridge density, indicating micro-morphological divergence at the varietal level. The R. mucronulatum complex is placed in the subgenus Rhododendron, which also includes R. lapponicum subsp. parvifolium and R. micranthum (Mo et al., 2022). Within the subgenus Hymenanthes, R. brachycarpum and R. aureum show a convergent seed morphology characterized by a striate surface, smooth periclinal wall, and medium to high ridge density with a smooth texture. However, when these micro-morphological patterns are assessed within the context of the established infrageneric classification, they do not consistently align with subgeneric boundaries (Wang et al., 2007; Szkudlarz, 2009). Specifically, seed micromorphological features are not supported by traditional classification systems; therefore, cluster analysis based on seed micromorphology does not endorse the current classification of the Rhododendron genus (Szkudlarz, 2009). Instead, seed coat traits reveal fine-scale differences among closely related taxa while also exposing morphological convergence across distantly related lineages. This suggests that micromorphology reflects a combination of lineage-specific signals and recurrent evolutionary patterns, rather than the hierarchical structure of the current classification (Fagúndez and Izco, 2004; Szkudlarz, 2009). As a result, while these seed features are not reliable diagnostic markers at the subgeneric level, they are useful for species-level identification within a given subgenus (Mo et al., 2022).
Although limited to ten taxa, this study provides new insights into the seed morphology of East Asian Rhododendron species. Their extremely small seeds made it difficult to detect macroscopic differences between the species. However, by combining traits such as color, size, weight, shape, and surface micro-morphology, species-level identification became possible. These findings emphasize the importance of evaluating combined morphological traits rather than relying on individual characteristics for accurate taxonomic classification. Nonetheless, given that this study mainly focused on external traits, it did not examine internal seed structures or physiological properties, which is a limitation. Future research should adopt a multidimensional approach, combining morphological data with seed anatomy, germination physiology, and functional traits to gain a more comprehensive understanding of seed evolution and taxonomy in Rhododendron.








