Taxonomic relationships of the subgenus Iris
In NJ tree based on ITS sequences (Fig. 1), the Subgen. Iris did not form a large branch but is divided into two branches (Clade A1, A2). Zhao (1985) classified this subgenus into two sections: Sect. Iris and Sect. Hexapogon. I. germanica in Sect. Iris clustered with I. pandurata and I. mandshurica in Sect. Hexapogon, while the others in Sect. Hexapogon grouped together. This is largely consistent with Wilson's (2017) research results, suggesting that there may be significant differences in the degree and timing of differentiation among species within the subgenus, indicating that the phylogenetic relationships within this subgenus need further study. I. goniocarpa var. tenella and I. goniocarpa formed a sister clade (99%), while I. goniocarpa var. grossa grouped with I. leptophylla, indicating that the relationship between I. goniocarpa and I. goniocarpa var. tenella was very close, whereas I. goniocarpa var. grossa was more distantly related to both. Observations reveal that the I. goniocarpa is only slightly more robust than I. goniocarpa var. tenella, with no significant differences in other aspects. I. goniocarpa var. grossa has a purplish-red flower color and the spots around the orange-yellow crest-like ornament on the outer tepals are larger and dark purple. The plant itself is larger and more erect, with broader and thicker leaves, distinctly different from the other two. Zhao et al. (2000) reclassified the I. goniocarpa var. grossa as I. cuniculiformis Noltie & K. Y. Guan, elevating it from a variety to a species. This study supports treating the I. goniocarpa var. grossa as a species and considers the I. goniocarpa var. tenella as a synonym of the I. goniocarpa.
In the gene tree constructed based on matK sequences (Fig. 3), all bearded irises clustered together (Clade D). I. pallida and I. germanica in the Sect. Iris formed a sister clade; both have rhizomatous roots and their seeds dehisce from the top of the fruit upon maturation. However, I. scariosa from the Sect. Hexapogon clustered with them, indicating a close phylogenetic relationship among the three. This is consistent with the findings of Zhong (2010). He speculated that this clustering result is due to their origin all in Central Asia and Europe and, they have identical chromosome number of x = 24. The remaining irises in the Sect. Hexapogon were divided into several clades, showing more complex phylogenetic relationships within the subgenus. Based on the results from both single-gene and multi-gene studies, Zhong (2010) also suggested that the subgenus classification is not uniformand. The phylogenetic relationships within the Subgen. Iris are complex and the division requires further study.
Taxonomic Relationships of the subgenus Crossiris
Subgen. Crossiris has 9 species and 2 varieties in China. In the ITS sequence phylogenetic tree (Fig. 1) clustering analysis, I. speculatrix was embedded in the Subgen. Limniris and formed a separate clade (Clade D), which is consistent with the results of molecular biology studies by Zhong (2010). Zhong suggested classifing the I. speculatrix into the Subgen. Limniris. Wu and Culter (2010) found that the leaf anatomical structure of I. speculatrix is similar to that of the Subgen. Limniris. Therefore, it is evident that its affinity with the Subgen. Limniris is relatively close. However, considering the characteristics of its sepal which has cristate ornaments, we do not support Zhong’s proposal to move this species into the Subgen. Limniris. Instead, we believe this species may be a transitional form between the Subgen. Limniris and the Subgen. Crossiris.
In the NJ tree based on the matK sequence (Fig. 3), except for I. proantha, for which no relevant sequence was collected, sequences for other species were collected. However, we collected the matK sequence for its variety I. proantha va,r. valida. Most species within the subgenus form a clade (Clade B2), but I. speculatrix, I. proantha var. valida, I. latistyla and I. tectorum are not clustered within this subgenus, indicating that this subgenus is not a natural taxon. I. speculatrix was embedded within the Subgen. Limniris and it forms a distinct clade (Clade B1) (73%), which is similar to the clustering result of this ITS analysis. I. latistyla is grouped into the subgen. Nepalensis (88%). This is consistent with the clustering results based on five chloroplast genes by Guo and Wilson (2013), who observed that the flower color and crest-like sepal ornament of this species are very similar to those of I. decora and I. collettii. Our study indicates that it is closely related to the subgen. Nepalensis, but its specific taxonomic position requires further data for exploration.
Taxonomic relationships of the subgenus Limniris
Zhao et al. (1985, 2000) divided species of the subgen. Limniris into three sections sect. Ioniris, sect. Limniris and sect. Ophioris. Our research also found that the subgen. Limniris exhibits multiple origins, as some elements of the sect. Limniris are located outside the core clade of the subgen. Limniris. In the ITS sequence NJ tree (Fig. 1), all species within this subgenus collected for this study cluster together (Clade D), but I. songarica is situated at the very edge of the subgenus, indicating it is more distantly related to the other species. I. ruthenica var. nana, I. lactea and its two varieties, as well as the I. halophila within the subgen. Xyridion, form a subclade. They are all relatively drought-tolerant or salt-alkali resistant. Zhao et al. (1985, 2000) classified I. lactea as the original species, with I. lactea var. chinensis and I. lactea var. chrysantha as varieties. However, Mathew (1989) argued that the floral color variations within this group are due to natural hybridization and that varieties should not be distinguished. Gao (1985) believed that I. lactea and I. lactea var. chrysantha are synonyms for I. lactea. In the NJ tree, this group clustered into a single clade(96%), indicating close relationships among the three. However, I. lactea and I. lactea var. chrysantha formed a sister clade(64%) and I. halophila was also embedded within this group, suggesting that the group is not a monophyletic origin. Our research results align with previous studies, supporting Zhao’s classification of I. lactea and I. lactea var. chrysantha as varieties of I. lactea as reasonable (Zhao et al. 2000).
Based on the matK sequence analysis, in the NJ tree (Fig. 3), the subgen. Limniris does not form a large clade but is instead divided into three clades (Clade A1, A2, A3), with species from other subgenera interspersed in between. This indicates that the subgenus is not a monophyletic group. The result was the same as that of in ITS-constructed NJ tree, I. halophila and its varieties are interspersed, forming a subclade with I. songarica and I. qinghainica and are clustered together with the sister clades of I. tenuifolia and I. loczyi (Clade A1). This is because these irises all grow in the relatively arid northwest region. I. odaesanensis, I. henryi, I. minutoaurea and I. rossii form a clade (Clade A2), generally consistent with Park's clustering based on the whole chloroplast genomes (Park et al. 2022). With the exception of I. henryi, all other species in this clade are native to the northeastern region. This clade also includes I. proantha var. valida from Subgen. Crossiris. Wilson (2009) believed that the phylogeny of the Subgen. Crossiris shows a certain geographic aggregation.
Taxonomic relationships of the subgenus Xyridion
In the phylogenetic tree based on ITS sequences(Fig. 1), I. halophila is the only species within the subgen. Xyridion. Mathew and Waddick suggested that it belongs to the subgen. Limniris (Mathew 1989; Waddick and Zhao 1992). In the NJ tree, this species is embedded within the subgen. Limniris (Clade D). This is consistent with the clustering results of multiple genes combined by Zhong (2010) and Mou (2011) also found that I. halophila grouped together with species within the subgen. Limniris through AFLP markers. The main characteristic of this subgenus is that the petals resemble a violin, but the midrib of the outer tepals lacks appendages, which aligns with the characteristics of the subgen. Limniris.
In the NJ tree of the matK sequence (Fig. 3), I. halophila var. sogdiana and I. halophila have been clustered into the subgen. Limniris, along with I. songarica and I. qinghainica, forming a subclade (Clade A1). This clade shares the same geographic distribution and similar external morphology. For example, both the I. halophila and I. songarica have violin-shaped petals, though the former has a sturdier flower stalk. The main difference between I. halophila and its variety lies in flower color: the former is yellow, while the latter can be blue-purple or variegated, with the claws of the inner and outer tepals being yellow and the upper parts being blue-purple. Based on the feature of having no appendages on the outer tepals, our findings are consistent with previous studies (Mathew 1989; Waddick and Zhao 1992; Zhong 2010): we recommend removing the species’ status as a separate subgenus and incorporating it into the Subgen. Limniris and we support Zhao’s treatment of I. halophila var. sogdiana as a variety of I. halophila (Zhao 1985).
Taxonomic Relationships of the subgenus Nepalensis
In the NJ tree constructed from the ITS sequence (Fig. 1), I. collettii and I. decora of the subgen. Nepalensis formed a sister clade (Clade B) (100%), indicating a close phylogenetic relationship. Both species exhibit the typical characteristics of the subgenus: fleshy spindle-shaped roots. We agree with previous research findings (Zhong 2010; Jiang et al. 2017) that this subgenus is a relatively natural taxonomic group. In the matK sequence study, the subgen. Nepalensis not only includes I. decora and I. collettii but also encompasses I. barbatula and I. collettii var. acaulis, which all cluster together (Clade B2) with 97% bootstrap support. The two newly added species also exhibit fleshy, spindle-shaped roots, consistent with the typical characteristics of this subgenus (Zhao et al. 2000).
Taxonomic relationships of the subgenus Pardanthopsis
The subgen. Pardanthopsis contains only two species, I. dichotoma and I. subdichotoma Y. T. Zhao. However, the classification status of I. dichotoma has always been controversial. Mathew (1989) believed that I. dichotoma should not be included in the Iris L. and classified it in the Pardanthopsis Lenz. Most scholars such as Baker, Dykes, Lawrence and Rodionenko (Baker 1892; Dykes 1913; Lawrence 1953; Rodionenko 1987) believed that I. dichotoma should be included in the Iris L. and placed it in the Subgen. Pardanthopsis. In the phylogenetic tree of the matK sequences, I. dichotoma was nested in the Iris L., indicating that I. dichotoma is closely related to the Iris L.. I. dichotoma together with Belamcanda chinensis, forms a separate clade, consistent with Wilson's clustering based on different genes and Feng's clustering based on complete chloroplast genomes (Wilson 2011; Feng et al. 2022). Comprehensive studies suggest that the classification status of the subgen. Pardanthopsis is reasonable. However, in the NJ tree (Fig. 3), I. subdichotoma is embedded in the subgen. Nepalensis, forming a sister clade with I. barbatula(66%). Zhong (2010) also found that this species clusters within the subgen. Nepalensis through multi-genome analysis. Based on palynological research, Qi and Zhao (1987) discovered that the pollen of I. subdichotoma is nearly spherical with a colpus of two-sulcus type, the exine stratification is not significant and it has a coarse reticulate ornamentation with granules within the meshes. These characteristics are similar to the pollen morphology of the subgen. Nepalensis and clearly distinguish it from the pollen morphology of the Subgen. Pardanthopsis: nearly spherical, single-sulcus colpus, distinctly two-layered exine, reticulate ornamentationand no granules within the meshes. Combining previous studies, we suggest transferring I. subdichotoma from the Subgen. Pardanthopsis to the Subgen. Nepalensis.
Classification of Belamcanda chinensis
In the NJ tree constructed from the ITS sequence (Fig. 1), Belamcanda chinensis was embedded within the genus Iris (Clade B) (70%). Baker, Dykes, Lawrence, Rodionenko and Mathew considered Belamcanda chinensis should be classified under the genus Belamcanda (Baker 1892; Dykes 1913; Lawrence 1953; Rodionenko 1987; Mathew 1989). However, Tillie, Wilson, Goldblatt and Mabberley argued that it should be classified under the genus Iris based on molecular data and external morphological characteristics (Tillie et al. 2000; Wilson 2004; Goldblatt and Mabberley 2005; Kang et al. 2020). Goldblatt and Mabberley even renamed the Belamcanda chinensis to Iris domestica (Goldblatt and Mabberley 2005). In the matK phylogenetic tree (Fig. 3), Belamcanda chinensis was embedded within the genus Iris and formed a single clade with I. dichotoma, indicating a closet relatives between the two. We suggested that Belamcanda chinensis be included in the Subgen. Pardanthopsis within the genus Iris.