We recorded morphological characters from the type specimen of Microrhamnus taquetii (i.e., A00003587) (Fig. 1A; Suppl. material 1: table SS1). Using these characters, we then searched for additional specimens exhibiting similar morphology through the online databases of the Global Biodiversity Information Facility (GBIF: https://www.gbif.org/) and the Chinese Virtual Herbarium (CVH: http://www.cvh.ac.cn/en). Subsequently, we compared these specimens with those of M. taquetii to determine the genus and species to which they should be assigned.

We reviewed the specimen records of the R. franguloides complex available through CVH and GBIF (Suppl. material 1: tables S1–S3). Based on these databases and herbaria, we compiled locality data and generated a distribution map (Fig. 2). When labels provided precise coordinates, these were transcribed directly; otherwise, coordinates were georeferenced from locality descriptions using online mapping resources (Suppl. material 1: table SS2). We subsequently conducted field surveys to collect fresh leaves and voucher specimens from representative populations for molecular and morphological analyses, respectively. The freshly collected leaves were immediately desiccated in silica gel. For each population, we obtained at least one representative voucher specimen for morphological analysis and sampled between two and 13 individuals for DNA sequencing (Table 1). In addition, we collected leaf samples from R. julianae and R. martini for sequencing. Geographic coordinates of the newly sampled populations were recorded in the field using a handheld global positioning system (GPS) device. All voucher specimens were deposited in the Herbarium of Xinjiang University (XJU).

We first examined morphological variation within the R. franguloides species complex. Morphological data were compiled from specimens accessed through GBIF and CVH following the methods of Lu et al. (2024). All quantitative traits were measured using ImageJ v. 1.49v, whereas qualitative traits were scored based on direct observations of the specimens. Newly collected and herbarium specimens were measured and compared within and among populations. At a larger scale, the specimens were divided into four geographic regions: China, South Korea, Japan, and the Liuchiu (Ryukyu) Islands (Figs 1, 2; Table 1), and morphological variation was further compared within and among regions. This approach enabled us to identify morphological variation both within and among populations or geographic regions. Morphological traits of leaves and branchlets were measured following the standardized protocols presented in Suppl. material 1: fig. S1, Tables 2, 3. The measured specimens of the R. franguloides complex covered its entire geographic range (Fig. 2). The type specimens of all synonyms were also examined (Fig. 1). Specifically, we measured seven specimens that were consistent with the protologue of R. franguloides var. inaequilatera (Ohwi 1951). To ensure consistency across the morphological dataset from all examined specimens, we selected seven quantitative and two qualitative traits of leaves and branchlets that exhibited clear differentiation between populations, geographic regions or taxa for the first morphological clustering analysis (Table 2). For example, lenticel density on old branchlet showed clear differentiation between populations, geographic regions or taxa, but we only obtained this dataset from 181 individuals in this complex, and all such traits were reflected by morphological comparison (Table 4). Our observations indicated that petiole length was relatively stable in leaves from the middle portions of branchlets throughout development. Therefore, only leaves from these portions were measured. Finally, a total of 412 specimens (388 identified as R. franguloides, including its varieties and synonyms, and 24 identified as R. caudata) were included in the first morphological clustering analysis, without missing data.

Fruit size is an important taxonomic character for species delimitation in Rhamnella (Lu and Sun 2019, 2020). Field observations indicated that fruit size in R. franguloides stabilized by July. Therefore, specimens collected before July and those without fruits were excluded from the second morphological analysis. According to the above operation procedure, we then retained 372 specimens of R. franguloides complex with stable fruit size for the subsequent clustering analysis based on all 13 quantitative and two qualitative traits (Table 3), without missing data. This dataset included most of the key traits traditionally used to distinguish species in Rhamnella, such as pubescence, petiole length, leaf length-to-width ratio, and dried fruit shape and size (Chen and Schirarend 2007; Lu and Sun 2019, 2020). The dataset comprised 325 specimens from China previously identified as Rhamnella franguloides (including type specimens of R. obovalis and B. congesta), 20 R. caudata specimens from Guangdong, and 27 R. franguloides specimens from the Liuchiu (Ryukyu) Islands, Japan, and Korea.

The above two morphological clustering analyses were analyzed using principal component analysis (PCA). For each dataset, PCA was performed in R version 4.4.3 (R Core Team 2025) using the prcomp() function, with all variables centered and scaled to unit variance (scale. = TRUE) so that traits measured on different scales contributed equally. Because each dataset included continuous, binary and ratio variables, a Gower dissimilarity matrix was computed (Maechler et al. 2026). After generating the PCA plots, we compared the morphological differences among the identified clusters by PCA plotting.

Total genomic DNA was extracted from silica-dried leaf material using a modified CTAB protocol (Doyle and Doyle 1990). Because the nuclear internal transcribed spacer (ITS) region has high discriminatory power at the species level in Rhamnella (Lu and Sun 2020), we preferentially selected this barcode for genetic delimitation. The ITS region was amplified using the primers ITS1a and ITS4a, according to the protocols described by Zhou and An (2025). PCR amplification was performed in 50 μL reactions containing 2 μL DNA template, 5 μL 10× PCR buffer, 0.8 μL dNTPs (2.5 mM), 2 μL of each primer (10 μM), 0.4–0.5 μL rTaq polymerase, and ddH₂O. Amplification conditions were as follows: an initial denaturation at 94 °C for 4 min; 38 cycles of 94 °C (40 s), 60 °C (45 s), and 72 °C (2 min); and a final extension at 72 °C for 6 min. PCR products were verified by agarose gel electrophoresis, and products of sufficient quality were sequenced.

To illustrate that the complex is distinct from R. julianae and R. martini, we constructed a maximum likelihood (ML) tree based on the ITS sequences using MEGA12 with default parameters (Kumar et al. 2024). Bootstrap analysis was performed with 1,000 replicates. The tree was rooted with Berchemiella berchemiifolia and B. wilsonii (Fig. 3). The phylogenetic analysis supported the distinction of the complex from R. julianae and R. martini (Fig. 3). Subsequently, we constructed a second ML tree using the same method to present the evolutionary relationship among the identified clusters. The second tree was rooted with two individuals of Rhamnella martini (X.L. Zhou 2025XS01 and X.L. Zhou 2025XS02). R. martini was selected as the outgroup because it is phylogenetically distinct from the R. franguloides complex (Fig. 3; Lu and Sun 2020).

Based on the type specimen of Microrhamnus taquetii (barcode No. A00003587), we extracted the following diagnostic morphological characteristics: a woody plant with simple, opposite leaves, ovate-elliptic to obovate, 2.0–5.5 cm long and 1–2.5 cm wide, with finely serrate margins, acuminate apices, and cuneate to obtuse bases; seven to eight pairs of lateral veins; petioles 1.0–3.0 mm long; branches opposite and distinctly bearing at least two rows of corky wings; branchlets subquadrangular; winter buds rounded, about 2 mm long, with the margins of the bud scales irregularly and sharply denticulate; fruits with at least 1–2 seeds, containing at least 1–2 locules, with a persistent sepal whose diameter is greater than that of the seed base (Fig. 1A). These characteristics, including opposite leaves and branches, branches with at least two rows of corky wings, subquadrangular branchlets, winter buds rounded and about 2 mm long, winter buds with bud scale margins irregularly and sharply denticulate, and fruit with a persistent sepal whose diameter is greater than the base of the seed, are typical characteristics of Euonymus alatus (a species of Euonymus in the family Celastraceae) (Fig. 1A–C). Therefore, M. taquetii should be excluded from the R. franguloides complex.

Two groups were visually recovered from the first PCA result (Fig. 4A). One (Cluster A) comprised all specimens from China previously treated as R. franguloides (including the type specimens of R. obovalis and Berchemia congesta) plus one specimen from South Korea (Fig. 4A; Suppl. material 1: tables S1–S3), another comprised R. caudata together with all R. franguloides specimens from Japan, the Liuchiu (Ryukyu) Islands and the remaining specimens from South Korea. These two groups were mainly distinguishable by the absence or presence of petiole and annual branchlet pubescence, petiole length, and leaf apex length (Table 4). The second PCA result, based on 372 fruiting specimens, recovered three major clusters (Fig. 4B). Cluster A remained unchanged, whereas the second group in Fig. 4A was further divided into Cluster B (all individuals of R. caudata) and Cluster C (27 specimens from Japan, the Liuchiu (Ryukyu) Islands and South Korea). Cluster B and Cluster C were mainly distinguished by lenticel density and dry fruit size (Table 4). Notably, the type specimen of R. japonica (L0553944) lacked fruits and was therefore not included in the second PCA result. However, its specific lenticel count (74 per square meter) and extremely low number of lateral veins (6–7) for this specimen (L0553944) further supported its placement in Cluster C rather than Cluster B (Table 4). For the two varieties within Cluster C, morphological boundaries were redefined, with a distinct difference in leaf petiole length (Table 4). The geographic distributions of the three major clusters (A, B, and C) and two varieties (var. inaequilatera and var. franguloides) within Cluster C were outlined (Fig. 5).

This study generated new ITS sequences from 34 individuals, comprising 29 individuals of the R. franguloides complex, two individuals of R. martini (PZ530792–PZ530793), and three individuals of R. julianae (PZ841209–PZ841211) (Tables 5 and Suppl. material 1: table SS4). In addition, four sequences of the complex were retrieved from GenBank (Fig. 5; Table 5): MH808997 (China, Qinling Mountains) and JF980328 (South Korea) in Cluster A, and KR083080 and JN900306 (Japan or the Liuchiu (Ryukyu) Islands) in Cluster C. Thus, a total of 33 sequences were analyzed for the complex. Eight unique ITS sequence types (PZ530784–PZ530791) were detected among the newly generated sequences from the R. franguloides complex (Table 5 and Suppl. material 1: table SS4). The three major clusters showed significant differences from one another. Cluster A was distinguished from the other two clusters (B and C) by fixed mutation sites at positions 14 (C), 45 (G), 46 (A), 49 (T), and 176 (T), and Cluster B exhibited such diagnostic nucleotides (i.e., fixed mutation sites compared with the other two clusters) at positions 29 (T), 52 (T), 173 (C), 179 (T), and 416 (G). In addition, Cluster C had nine private sites when compared with either Cluster A or Cluster B, whereas Cluster B had ten private sites when compared with either Cluster A or Cluster C.

Molecular phylogenetic analysis based on ITS sequence variation showed that the R. franguloides complex was reciprocally monophyletic with respect to R. julianae and R. martini (Fig. 3). Within the complex, Cluster A formed an independent clade that was sister to a well-supported clade containing Cluster B (R. caudata) and Cluster C (Fig. 6). Within the clade comprising Clusters B and C, the R. caudata (i.e., Cluster B) subclade comprised 13 individuals from Guangdong, China, whereas two individuals of Cluster C, representing both var. inaequilatera and var. franguloides or either of them, formed another subclade (Fig. 6; Table 1 and Suppl. material 1: table SS4). These two subclades were also well supported.

Kim (1993) and Ma (2001) treated Microrhamnus taquetii (the type specimen voucher No.: T. Taquet 153) as a synonym of Euonymus alatus, consistent with the annotation by Kim (February 1993) on the type sheet (barcode No. A00003587; Fig. 1A). However, Chen and Schirarend (2007) argued that M. taquetii should instead be treated as a synonym of R. franguloides (Ma and Funston 2008; IPNI 2026; POWO 2026). Our direct re-examination of the type specimen (barcode No. A00003587) supports their treatments and the annotation by Kim (February 1993) on the type sheet, because the specimen exhibits a set of characters that are entirely inconsistent with Rhamnella but perfectly match those of Euonymus alatus (Celastraceae) (Fig. 1; Chen and Schirarend 2007; Ma and Funston 2008). These characters include opposite leaves and branches, branches bearing at least two rows of corky wings, subquadrangular branchlets, and rounded winter buds with irregularly and sharply denticulate bud-scale margins collectively, these are key characters for identifying Euonymus alatus (Ma and Funston 2008). In addition, they are fundamentally inconsistent with those of R. franguloides and its allies (Chen and Schirarend 2007). The type specimen of M. taquetii was collected in South Korea, where E. alatus also occurs (Fig. 1A–C). Therefore, the treatment of M. taquetii as a synonym of R. franguloides should be rejected, whereas its treatment as a synonym of Euonymus alatus, as proposed by Kim (1993), the annotation by Kim (February 1993) on the type sheet, and Ma (2001), is supported.

Previous taxonomic treatments of the R. franguloides complex commonly recognized two species, one of which comprised two varieties: R. franguloides (var. inaequilatera and var. franguloides) and R. caudata (Ohwi 1951; Chen and Schirarend 2007; IPNI 2026; POWO 2026). Several names were also treated as synonyms of R. franguloides (Figs 1, 2), and no synonym was treated under R. caudata (Chen and Schirarend 2007). Population-level morphological clustering and comparison, together with geographic distributions, supported the recognition of at least three distinct clusters (A, B, and C) in this complex (Figs 4, 5; Table 4). Cluster A corresponds to the specimens previously identified as R. franguloides (including type specimens of R. obovalis and B. congesta) in China and one specimen from Korea, while Cluster B comprises only specimens of R. caudata from Guangdong, and Cluster C includes all specimens of the two varieties from Japan and the Liuchiu (Ryukyu) Islands and the remaining specimens from South Korea (Fig. 5; Table 4). These three clusters can be distinguished from one another based on the absence or presence of petiole and annual branchlet pubescence, lenticel density, petiole length, leaf apex length, and dry fruit size in the complex (Table 4). This finding differs markedly from the previous description and treatment (Chen and Schirarend 2007). Within Cluster C, var. inaequilatera was previously distinguished from var. franguloides by unequal-sided leaf bases, two–locular pyrenes, striate fruit surfaces, and leaf petiole length (10–15 mm) (Ohwi 1951). Among them, only leaf petiole length (3.8–7.6 mm vs. 7.8–15.8 mm) can distinguish the two varieties (Table 4), with new variable range different from the previous description (Ohwi 1951).

Allopatric speciation is considered a common model through which reproductive isolation develops (Coyne and Orr 2004). Although the development of reproductive isolation and morphological discontinuities in allopatry may not be fully synchronized (Wu 2001; De Queiroz 2007; Liu 2016), most units exhibiting morphological discontinuities may show varying degrees of reproductive isolation (Mallet 2008). In addition, geographical discontinuity is itself an important isolating mechanism that maintains morphological and genetic boundaries (Schluter 2001). In this study, Cluster B (R. caudata) is restricted to southern China (Guangdong), whereas Cluster C is distributed in South Korea, the Liuchiu (Ryukyu) Islands, and Japan (Fig. 5). Clearly, these two distinct clusters are geographically discontinuous, and no evidence suggests that the two widely recognized species should be merged into a single species. Therefore, the recognition of Cluster B (corresponding to R. caudata) and Cluster C (corresponding to R. franguloides and its synonym from Japan) as separate species should remain unchanged (Figs 1, 2, 3, 4, 5; Table 4). Furthermore, Cluster A differs considerably in morphology from Clusters B and C (Fig. 4; Table 4), with B and C hypothesized to be the most closely related pair, and Cluster A is also largely or completely allopatric with respect to both (Fig. 5). The Cluster A was previously described as R. obovalis or B. congesta. Berchemia congesta (1875) predates R. obovalis (1914) (Chen and Schirarend 2007). Under the rules of nomenclature, Rhamnella congesta should be the legitimate name for Cluster A, with B. congesta as the homotypic synonym and R. obovalis as its heterotypic synonym. Based on the morphological evidence and geographical distributions, these three major clusters identified here correspond well to R. congesta, R. caudata, and R. franguloides. The two varieties (var. franguloides and var. inaequilatera) of the newly-delimited R. franguloides can be distinguished only by leaf length (Table 4). Although they exhibit pronounced geographical separation (Fig. 5), there is no other new evidence to support the elevation of var. inaequilatera to species rank. Thus, the existing taxonomic treatment of the two varieties is retained.

Overall, our morphological delimitation of this complex differs markedly from previous taxonomic treatments in terms of taxon descriptions, circumscriptions and geographical distributions, highlighting the importance of population‐level analyses of morphological variation for delimiting species boundaries and informing taxonomic decisions.

Molecular phylogenetic analysis is one of the most important approaches for inferring evolutionary relationships and delimiting species (e.g., Hu et al. 2015; Lu et al. 2024). Morphological evidence indicated that R. caudata and the newly-delimited R. franguloides were likely the most closely related pair in this complex. However, it is also necessary to identify the key morphological characters that reflect phylogenetic relationships. Our sequence alignment revealed many variable sites that clearly distinguish the three species (Table 5). Either R. congesta (Cluster A) or R. caudata (Cluster B) possessed five fixed diagnostic nucleotide sites compared to the other two species. In addition, R. franguloides (Cluster C) possessed nine private sites that distinguished it from either Cluster A or Cluster B, whereas Cluster B had ten private sites when compared with either Cluster A or Cluster C (Table 5). These findings suggested strong genetic divergence among these three clusters or species. Therefore, R. caudata (Cluster B) is one well-delimited species different from R. franguloides (Cluster C), by integrating multiple lines of evidence from morphology, geographic distribution, and genetic divergence. Furthermore, the phylogenetic tree showed that R. congesta (Cluster A) forms an independent clade sister to the well-supported clade containing R. caudata (Cluster B) and R. franguloides (Cluster C) (Fig. 6). This finding suggested that the absence or presence of petiole and annual branchlet pubescence, petiole length, and leaf apex length are congruent with the inferred phylogenetic relationships (Figs 5, 6; Table 4). Under an integrative framework for species delimitation (Hu et al. 2015; Liu 2016; Lu et al. 2024), R. congesta (Cluster A) as a species was strongly supported by morphological and molecular evidence.

It should also be noted that leaf petiole length can well reflect the phylogenetic relationship (Table 4; Fig. 6). The variety R. franguloides var. inaequilatera was originally described as a species (Ohwi 1951), only with the longer leaf petiole (7.8–15.8 mm) than var. franguloides (3.8–7.6 mm) (as redefined in this study, Table 4). Although morphological difference among species in Rhamnaceae are often subtle (Chen and Schirarend 2007), whether this variety should be reinstated at species rank remains unclear. If both ITS sequences are assigned to either variety, it is likely that var. franguloides is more closely related to R. caudata. The acceptance of this hypothesis would imply that R. inaequilatera should be resurrected. However, the present study cannot determine whether the downloaded ITS sequences of R. franguloides should be assigned to both varieties or either variety (Table 5). To resolve this, more samples of R. franguloides should be collected for molecular phylogenetic and morphological analyses. Additionally, the present study relied solely on ITS marker, and sampling of Cluster C remains limited. Further studies incorporating additional molecular markers and expanded sampling could help further clarify relationships and genetic divergences within the R. franguloides complex.

The following key is based on diagnostic characters identified from all specimens examined in the present study. These three species are most reliably distinguished by combinations of these characters. For the two varieties within R. franguloides, single leaf petiole length can separate them completely. For taxa with more than two characters listed in the key, we strongly recommend using all of them for field identification.