Fifty-three experimental taxa, including 51 Epimedium taxa and two Vancouveria species (the most closely-related genus) were collected from China, Japan, Germany and the United States. The details, including collection location and voucher specimens of these experimental materials, are detailed in Table 1. All the voucher specimens were conserved in the Institute of Karst Research, Guizhou Normal University. Parts of materials also were cultivated in a greenhouse of the Institute.

The preparation method of chromosome slides referred to the method of Sheng and Chen (2007a). Chromosome slides obtained were examined and photographed by a tri-ocular microscope with a CCD camera (BX52-DP72, Olympus, Japan). At least ten well-spread metaphase plates of each experimental taxa were selected to calculate the karyotype parameters. Long arm (L), short arm (S), total chromosome length (TCL), relative length (%), arm ratio (L/S), arm index and location of the centromere were measured and calculated. RLR (the relative length ratio of the longest and shortest chromosome) and P (the proportion of chromosomes with arm ratio over 2:1) were also calculated. The classification of chromosome type was conducted according to Levan et al. (1964) and the karyotype type was determined by Stebbins’ (1971) criteria.

The calculation formula of karyotype similarity coefficients referred to Tan and Wu (1993):

$\ddot{\mathrm{e}}=\tilde{\mathrm{a}}\times \beta$ (1)

${\tilde{a}}_{ij}=\frac{{\sum }_{k=1}^n{x}_{ik}·x_{jk}-\left({\sum }_{k=1}^n{x}_{ik}\right)\left({\sum }_{k=1}^n{x}_{jk}\right)}{\sqrt{{\sum }_{k=1}^n{\left(x_{ik}-\overline{x}\right)}^2}·\sum {\left(x_{jk}-{\overline{x}}_j\right)}^2}$ (2)

$\beta =1-\frac{d}{D}$ (3)

In formula (3), D is the sum of distances and d is the square root of the product of inner distance (d_i) and outer distance (d_n):

$D=\sum _{k=1}^n⌊x_{ik}⌋+\sum _{k=1}^n⌊x_{jk}⌋$ (4)

$d=\sqrt{d_i·d_n}$ (5)

$d_i=\sum _{k=1}^n\left|x_{ik}-x_{jk}\right|$ (6)

$d_n=\left|\sum _{k=1}^n\left|x_{ik}\right|-\sum _{k=1}^n\left|x_{jk}\right|\right|$ (7)

According to the above formulae, the karyotype similarity coefficients of mitotic metaphase chromosomes of all the 53 experimental taxa were calculated. Furthermore, the 53 experimental taxa were clustered, based on the karyotype similarity coefficients by the UPGMA method (Tan and Wu 1993). The cluster analysis was conducted and the dendrogram was drawn by the NTSYS-pc software (Version 2.10e).

All 51 studied Epimedium taxa are diploid with 12 chromosomes (2n = 2x = 12). Pair I of homologous chromosomes in each studied taxon has one pair of satellites. The mitotic metaphase chromosomes of the 51 Epimedium taxa studied were illustrated in Fig. 1 (Material 1–24) and Fig. 2 (Material 25–51), respectively. The karyotype parameters of chromosome length, arm ratio, centromere index, RLR, P and karyotype are all detailed in Table 2. The average lengths of the six pairs of homologous chromosomes were 9.82, 9.04, 8.59, 8.02, 7.56 and 6.98 μm, respectively. The average arm ratios of the six pairs of homologous chromosomes were 1.28, 1.31, 1.44, 1.98, 2.12 and 2.06, respectively. In all 51 Epimedium taxa studied, the majority of chromosomes were metacentric (m) chromosomes or submetacentric (sm) chromosomes. Subtelocentric (st) chromosomes were few and telocentric (t) chromosomes were not found. The average values of RLR and P of the 51 taxa studied were 1.41 and 0.33, respectively. According to Stebbins’ (1971) classification, the karyotypes of all Epimedium taxa tested were highly symmetrical with type 2A or 1A. It also can be seen that the karyotypes between species were very similar in Epimedium species.

Figure 1. 

Mitotic metaphase chromosomes in 24 Epimedium species. 1 E. ecalcaratum 2 E. shuichengense 3 E. platypetalum 4 E. davidii 5 E. pauciflorum 6 E. flavum 7 E. ilicifolium 8 E. mikinorii 9 E. membranaceum 10 E. lishihchenii 11 E. acuminatum 12 E. wushanense 13 E. leptorrhizum 14 E. baojingense 15 E. chlorandrum 16 E. luodianense 17 E. pudingense 18 E. glandulosopilosum 19 E. pseudowushanense 20 E. franchetii 21 E. enshiense 22 E. sutchuenense 23 E. zhushanense 24 E. pubescens. Scale bars: 5 μm.

The two Vancouveria species studied also were diploid with 12 chromosomes (2n = 2x = 12). Furthermore, pair I of homologous chromosomes of each species also has one pair of satellites located. The mitotic metaphase chromosomes and the karyotype parameters of the two species studied are given in Fig. 2 (Material 52, 53) and Table 2 (Material 52, 53), respectively. The average lengths of the six pairs of homologous chromosomes were 9.65, 9.11, 8.69, 8.11, 7.30 and 7.16 μm, respectively. Furthermore, the average arm ratios of the six pairs of homologous chromosomes were 1.17, 1.10, 1.19, 1.43, 1.93 and 1.59, respectively. In the two Vancouveria species studied, only two chromosomal types, i.e. m and sm types, were found. The average RLR of the two species studied was 1.41. The P values of the two species studied were 0.00 and 0.17, respectively. Moreover, the karyotypes of the two species studied were 1A and 2A, respectively. It also can be seen that the karyotypes of the two Vancouveria species were very similar.

Figure 2. 

Mitotic metaphase chromosomes in 27 Epimedium taxa and two Vancouveria species. 25 E. sagittatum 26 E. sagittatum var. glabratum 27 E. dolichostemon 28 E. truncatum 29 E. brevicornu 30 E. myrianthum 31 E. stellulatum 32 E. fargesii 33 E. elachyphyllum 34 E. koreanum 35 E. grandiflorum var. grandiflorum 36 E. grandiflorum var. thunbergianum 37 E. grandiflorum var. higoense 38 E. grandiflorum var. coelestre 39 E. sempervirens 40 E. sempervirens var. hypoglaucum 41 E. sempervirens var. multifoliolatum 42 E. trifoliatobinatum 43 E. diphyllum 44 E. cremeum 45 E. kitamuranum 46 E. setosum 47 E. alpinum 48 E. pubigerum 49 E. pinnatum subsp. colchicum 50 E. pinnatum cv. “Elegans” 51 E. perralderianum 52 V. hexandra 53 V. chrysantha. Scale bars: 5 μm.

The karyotype similarity coefficients of the 53 taxa studied were calculated (Suppl. material 1: Table S1). The 53 taxa are clustered into two groups, mostly corresponding to the two genera (Fig. 3). However, some species of Epimedium, i.e. E. truncatum, E. leptorrhizum and E. dolichostemon were clustered into the group of genus Vancouveria.

Figure 3. 

Diagram of cluster analysis of karyotype similarity coefficients in 51 Epimedium taxa and two Vancouveria species.

Clustering results also showed Epimedium, firstly, split into two groups, E. subg. Epimedium and E. subg. Rhizophyllum. Epimedium subg. Epimedium is further split into two clusters, basically reflecting the geographical distribution, with one group mainly consisting of the taxa native to the Mediterranean and West Asia and the majority of the species of the other cluster native to east Asia. The clustering result also showed that the genetic diversity of Epimedium taxa native to east Asia was higher than those of Mediterranean and West Asian taxa. Finally, some species native to east Asia, i.e. E. brevicornu, E. trifoliatobinatum and E. flavum, were clustered into the group of the Mediterranean and West Asian taxa.

The karyotype similarity coefficients between the original species and its variant were significant, such as E. sempervirens var. sempervirens and E. sempervirens var. hypoglaucum, E. sempervirens var. multifoliolatum, E. sagittatum var. sagittatum and E. sagittatum var. glabratum and E. grandiflorum var. grandiflorum, E. grandiflorum var. thunbergianum, E. grandiflorum var. higoense and E. grandiflorum var. coelestre. It can be seen that the cluster analysis of karyotype similarity coefficients can provide reliable clues for studies on plant taxonomy and systematics, especially for those taxa with similar karyotypes between species and insufficiency differences between homologous chromosomes.

Results of karyotype analysis showed that the genome of Epimedium was conservative in evolution and highly similar between species. The 51 Epimedium taxa tested are all diploid with the basic chromosomal number of 6 (2n = 2x = 12) and each of them has one pair of satellites located on pair I of homologous chromosomes. These results are consistent with the previous research reported (Tanaka and Takahashi 1981; Takahashi 1989; Yan et al. 2016; Zhang et al. 2018). Karyotypes of the 51 taxa studied are all highly symmetrical with the type 2A or 1A of Stebbins (1971). There has been the conclusion that species with symmetrical karyotypes usually are ancient and primitive in evolution and the karyotypes of evolutionary taxa are asymmetrical in spermatophytes (Stebbins 1971; Stace 2000). The highly symmetrical karyotypes in Epimedium species indicate that Epimedium should be conservative in evolution, consistent with other studies on the morphology (Stearn 2002; Xu et al. 2019), molecular biology (Kim and Jansen 1998; Kim et al. 2004), C-banding (Tanaka and Takahashi 1981; Takahashi 1989) and rDNA chromosomal location (Sheng and Wang 2010) in this genus. However, for a long time, studies on the systematics and taxonomy of Epimedium by using the traditional karyotype analysis have achieved little, because of the highly similar karyotypes between species and insufficient differences between homologous chromosomes (Guo et al. 2008).

Epimedium is phylogenetic related to Vancouveria (Tishler 1902; Berg 1972; Loconte and Estes 1989; Jin et al. 2018), being the species belonging to Vancouveria once classified into Epimedium (Stearn 1938). In the present study, results showed that the karyotypes of Epimedium and Vancouveria were highly similar. Karyotypes of the two Vancouveria species studies, with one pair of satellites located on pair I of homologous chromosomes, are very similar with some species of Epimedium. The two Vancouveria species, i.e. V. hexandra and V. chrysantha, were clustered into the group with E. truncatum, E. leptorrhizum and E. dolichostemon in the clustering of karyotype similarity coefficients. When compared with Diphylleia Michaux, Dysosma R. E. Woodson, Podophyllum L., and Sinopodophyllum Ying, the karyotype of Epimedium is significantly different and more symmetrical (Li 1986; Ma and Hu 1996), suggesting that the genus might be an ancient taxon in Berberidaceae and distantly related to these four genera. This conclusion also can be well supported by studies on morphology (Li et al. 2014), palynology (Zhang and Wang 1983; Wang et al. 2015), molecular markers (Wang et al. 2001; Sun et al. 2005; Zhang et al. 2014, 2016), isozymes (Sheng et al. 2011) and chemotaxonomy (Koga et al. 1991; Sheng et al. 2008). Therefore, it can be seen that karyotype analysis has important significance for studies on the relationships of Epimedium within Berberidaceae.

Cluster analysis of karyotype similarity coefficients showed that, although the genome was conservative in evolution and the karyotypes between species were highly similar, the cluster analysis of karyotype similarity coefficients still provided some valuable clues for studies on phylogenetics and taxonomy in Epimedium. Clustering results of karyotype similarity coefficients strongly support the classification of the two subgenera of E. subg. Rhizophyllum and E. subg. Epimedium. Based on morphological characteristics and geographical distribution, E. Subg. Epimedium is clustered into two groups: 1) Mediterranean and Western Asian taxa and 2) East Asian taxa. The present results supported this classification, consistent with the previous studies on the morphology (Stearn 2002), cytology (Sheng et al. 2010; Zhang et al. 2018), molecular biology (Kim et al. 2004; Guo et al. 2018; Sajad et al. 2018) and phytochemistry (Guo et al. 2008), all of which confirmed the conclusion that the systematic relationship between species is closely related to the specific geographical distribution in this genus. The present results also showed that, although the vast majority of East Asia taxa could be clustered into one group, there are a few species clustered into other groups of Epimedium or the group of Vancouveria, indicating that the genetic diversity of East Asian taxa are the highest in the genus.