Iris sanguinea is conspecific with I. sibirica (Iridaceae) according to morphology and plastid DNA sequence data

Morphological study

The I. subser. Sibiricae species descriptions available in literature (Krylov, 1929; Sergievskaya, 1972; Doronkin, 1987; Mathew, 1989; Pavlova, 1987; Zhao, Noltie & Mathew, 2000; Grey-Wilson, 2012) were examined. We evaluated the thirteen characters, which were selected from those typically used in the literature together with those considered relevant according to our personal observations (see Fig. 1). These characters are listed in detail in Table 1. The original material of I. sanguinea, I. sibirica, I. sibirica var. haematophylla, I. orientalis Thunb., and I. typhifolia (see Taxonomic treatment below) was studied. In total, 224 scaled specimens of well-developed plants in flowering or fruiting were measured (see Appendix S1). The specimens of I. sanguinea and I. sibirica have been checked through high resolution images available in virtual herbaria (herbarium codes according to Thiers, 2020): ABGI and VBGI (https://botsad.ru/herbarium/), E (https://data.rbge.org.uk/search/herbarium/), MHA and MW (https://plant.depo.msu.ru/), NS and NSK (http://herb.csbg.nsc.ru:8081/#fuzzy-label), PI, PRC and WU (https://herbarium.univie.ac.at/database/search.php). For I. typhifolia, 48 specimens were used: 27 specimens from the Chinese botanist Yu-Tang Zhao, an expert on Chinese Iridaceae (e.g., Waddick & Zhao, 1992; Zhao, Noltie & Mathew, 2000), collection at NENU, and also 21 specimens have been checked through images available in virtual Chinese herbaria ( http://www.cvh.ac.cn/). The morphological characters were measured using AxioVision 4.8 (Carl Zeiss, Germany), a freeware comprehensive images viewer.

For morphometric data analysis, nine characters were used (see Table 1). In this study both parametric and non-parametric versions of a one-way variance analysis (ANOVA) were applied. The differences were considered significant at p-value <  0.05. As multiple statistical testing was performed, the calculated p-value was adjusted using the procedure proposed by Benjamini & Hochberg (1995). To test basic ANOVA assumptions Shapiro–Wilk test for normality and Levene’s test for equality of variances were used. The missing values in the original data table were imputed using corresponding median values according (Kuhn & Johnson, 2018). The Kruskal–Wallis test was chosen as a non-parametric ANOVA algorithm (Dodge, 2008). Principal components analysis (PCA) was used to visualize the distribution of the analyzed individuals over the space of morphometric characters. It was applied to all quantitative characteristics. Directions of principal components were described in the factor space by their highest correlation values (denoted by r) with original axes. Computations were performed by means of SciPy (Virtanen et al., 2020) and Scikit-Learn (Pedregosa et al., 2011) packages.

DNA extraction, amplification and sequencing

Sequences of four cpDNA regions were obtained for 44 specimens taken from wild populations, herbarium material and living collections. Among those, there were 20 from 13 localities in the I. sibirica distribution range, 22 from 11 localities in the I. sanguinea distribution range, and two plants were of unknown origin (Fig. 2). It was not possible to obtain samples from Japan and northeastern China, including Liaoning Province, where I. typhifolia was described from. Nevertheless, while searching GenBank for any sequences of four studied cpDNA regions of the I. subser. Sibiricae species, we found sequences of only either psbA–trnH or trnL–trnF for several accessions of I. sibirica and I. typhifolia, as well as I. sanguinea from Japan, northeastern China and the Republic of Korea (see Table S1). The sequences of four cpDNA regions from the complete chloroplast genome of I. sanguinea from the Republic of Korea (Lee et al., 2017) were included in the study. Our sampling also comprises representatives of three other series of I. sect. Limniris: (1) I. laevigata Fisch., I. ensata Thunb., and I. pseudacorus L. from I. ser. Laevigatae (Diels) G.H.M.Lawr.; (2) I. lactea Pall., I. oxypetala Bunge, and I. tibetica from I. ser. Lacteae; (3) I. uniflora Pall. ex Link from I. ser. Ruthenicae (Diels) G.H.M.Lawr. Iris dichotoma Pall. from I. subgen. Pardanthopsis (Hance) Baker was used as outgroup. The complete specimen list, including the sampling localities and the voucher information is given in Table 2.

DNA extraction, amplification, and direct sequencing of four non-coding cpDNA regions (trnS–trnG, trnL–trnF, rps4–trnS^GGA, and psbA–trnH) follows Kozyrenko et al. (2004); Kozyrenko, Artyukova & Zhuravlev (2009). The cycle sequencing was accomplished on both strands and fragments were separated using a genetic analyzer ABI 3130 (Applied Biosystems, USA) in the Instrumental Centre of Biotechnology and Gene Engineering (Vladivostok, Russia). Sequences were deposited in the European Nucleotide Archive database; their accession numbers are available in Table 2.

Data analysis

The sequences of each cpDNA region obtained in this study and retrieved from the complete chloroplast sequence of I. sanguinea (KT626943) were aligned manually using the program SeaView v. 4 (Gouy, Guindon & Gascuel, 2010) and concatenated for each specimen. We included in the dataset indels and length variation in mononucleotide repeats because repeatability tests allowed us to exclude PCR errors. The haplotypes were identified based on combined DNA sequences using DnaSP v. 5 (Librado & Rozas, 2009). This program was also used to calculate the degree of divergence between cpDNA sequences based on nucleotide substitutions. A haplotype network was built using Network v. 4.6 (Bandelt, Forster & Röhl, 1999), treating each deletion/insertion, regardless of size as a single mutational event and using the median joining (MJ) algorithm with default settings. To reveal relationships between I. sanguinea, I. sibirica, and I. typhifolia, a haplotype network was also built using a dataset including psbA–trnH and trnL–trnF sequences obtained in our study and sequences of I. typhifolia retrieved from GenBank.

Phylogenetic analyses were performed on two datasets of combined sequences for four cpDNA regions studied (available at https://purl.org/phylo/treebase/phylows/study/TB2:S26635). The first one was composed of sequences from the I. subser. Sibiricae specimens obtained in the present study, haplotypes of seven taxa of I. sect. Limniris and I. dichotoma as outgroup. The second dataset was enlarged by the addition of psbA–trnH and/or trnL–trnF sequences for 13 accessions of the I. subser. Sibiricae species available in GenBank, and for these accessions, lacking portions of sequences (trnS–trn G and rps4–trnS regions) were coded as missing. Phylogenetic analyses were performed using Maximum Likelihood (ML) and Maximum Parsimony (MP) methods as implemented in PAUP v. 4.0b10 (Swofford, 2003). Bayesian Inference (BI) was conducted using MrBayes v.3.2.6 (Ronquist & Huelsenbeck, 2003) on the CIPRES portal (http://www.phylo.org/; Miller, Pfeiffer & Schwartz, 2010). For the MP analyses, gaps were coded according to (Simmons & Ochoterena, 2000), as implemented in the program FastGap v. 1.2 (Borchsenius, 2009). Optimal trees were found using a heuristic search with 1,000 random addition sequence replicates, starting trees obtained via stepwise addition, tree bisection and reconnection (TBR) branch swapping and the MulTrees option in effect. For ML and BI analyses, GTR + I + G model was selected according to the Akaike information criterion (AIC) using Modeltest v. 3.6 (Posada & Crandall, 1998). ML heuristic searches were done using the resulting model settings, 100 replicates of random sequence addition, TBR branch swapping and MULTrees option on. In BI, using the default prior settings, two parallel MCMC runs were carried out for ten million generations, sampling every 1,000 generations for a total of 10,000 samples. Convergence of the two chains was assessed, and the posterior probabilities (PP) were calculated from the trees sampled during the stationary phase. The robustness of nodes in ML and MP trees was tested using bootstrap with 1,000 replicates (bootstrap percentage, BP).

Morphological data

Morphological comparison among the I. subser. Sibiricae species is provided in Table 3. The results showed overlap of I. sanguinea, I. sibirica, and I. typhifolia at the morphological level (Fig. 3, Table 3). The majority of characters were variable in this analysis (see Coefficient of variation in Table S2).

The result of PCA revealed three characters with high factor loadings (r ≥ 0.5) on the first three principal components. These are LL, SH and CL (see abbreviations in Table 1). Together, the first three components accounted for 99.2% of the total variation. The first two components explained 75.3% and 21.9% of the total variation, respectively.

The biplot of PCA for all those species illustrates the overlap between all specimens and significant morphological similarity (Fig. 3). Two characteristics SH and LL displayed the highest correlations with the first and second axis (corresponding values are r = 0.73 and r = 0.67), and the remaining one (CL) highly influenced the third axis (r = 0.93). Results of parametric and non-parametric ANOVA analysis to projected data on three principal components showed that mean (median in case of the non-parametric test) values do not differ significantly among the species. Corresponding statistics and p-values are: p-value = 0.21 and adjusted p-value = 0.63 for traditional ANOVA; p-value = 0.03 and adjusted p-value = 0.11 for Kruskal–Wallis test. However, being applied to the original plant characters, both parametric and non-parametric ANOVA tests showed significant differences of average values for I sanquinea, I. sibirica, and I. typhifolia. Our results showed that mean (in case of traditional ANOVA) and median (in case of Kruskal–Wallis test) values only for LL and possibly PL do not significantly differ among the considered species (Table S2). Thus, having likely different average values of morphometric characters, caused by environmental conditions and interspecific trait variability, these species can still be considered as indistinguishable in a generalized (PCA) factor space.

Molecular data

Among the 44 specimens studied, nine haplotypes (H1–H9) were identified based on nucleotide substitutions and indels detected across 3766 aligned positions of four cpDNA regions (Table 2). Four haplotypes (H6–H9) were unique, i.e., found in a single population: H6 in RLV population (Leningrad Oblast, Russia), H7 in population RMS from the Setun River valley (Moscow Oblast, Russia), H8 in population GBB from Georgia, while H9 was found in the plant Sc1 cultivated at the Botanic Garden of Cambridge University, the United Kingdom (UK). Five other haplotypes were detected in more than one accession, often from geographically distant locations in the I. subser. Sibiricae distribution range. The sequences of cpDNA regions obtained in our study were compared with those from the complete chloroplast sequence of I. sanguinea from the Republic of Korea (KT626943). Haplotype H1 found in accessions from two localities in Russia (RP1, RKY) and from three localities in Mongolia (BAD, MDB, and MKB), turned out to be identical with the haplotype of I. sanguinea from the Republic of Korea (KT626943). Specimens of populations RP3, RP4, and RP5 from Primorsky Krai, Russia shared haplotype H3, while populations ORL, RP2 (Primorsky Krai) and RCH (Amur Oblast, Russia) shared haplotype H2. Specimens from populations ALS, ALT, and ZOR (Armenia), GJP (Georgia), RKT (Karachay-Cherkess Republic, Russia), RRU (Udmurt Republic, Russia), and RKP (Kurgan Oblast, Russia) shared haplotype H4. Haplotype H5 was found in samples RKU (Kaluga Oblast, Russia), RPS (Pskov Oblast, Russia), and LAS (Austria) as well as in a cultivated plant Sc2 (UK). No specimen from the European part of the distribution range shared haplotypes with plants from the Asian part. The sequence divergence of cpDNA between plants from the European and Asian parts of the distribution range was very low (K_S = 0.00056).

In the median network, all haplotypes formed one group (Fig. 4A) with a minimal divergence between each other (one to three mutational steps). Five haplotypes (H1–H3, H7, and H9) formed a star-like structure with haplotype H1 in the centre. This group composed of all haplotypes (H1–H3) from East Asian plants also included H7 from Eastern Europe and differed only by one substitution in the psbA–trnH region from all other haplotypes found in plants from the European range, namely haplotypes H4–H6 and H8. All haplotypes found across the I. subser. Sibiricae distribution range were closely related and derived from the same unsampled or extinct ancestral haplotype connected by many mutation steps with the haplotype of I. pseudacorus from I. ser. Laevigatae (Fig. 4A). A similar pattern was obtained in the network based on sequence data from the psbA–trnH and trnL–trnF regions, which included sequences of I. typhifolia retrieved from GenBank (Fig. 4B). In this network, all specimens from the Asian part of range share the common haplotype connected by six mutational steps with haplotype of I. typhifolia and by two steps with two haplotypes found in specimens from the European range.

MP, ML and BI analyses based on sequences of I. subser. Sibiricae obtained in the present study yielded similar topologies with few differences in node statistical supports (Fig. 5A). All Iris specimens clustered into highly supported (BP 100, 100%, PP 1.0) clades according to their affiliation to corresponding series of I. sect. Limniris. Haplotypes of all plants belonging to I. subser. Sibiricae formed a monophyletic highly supported clade (BP 100, 100%, PP 1.0) sister to the clade including species of I. ser. Laevigatae (BP 82, 93%, PP 1.0). Within the I. subser. Sibiricae clade, it was possible to distinguish a group including haplotypes H1–H3 from the Asian part of range, haplotype H7 from the Moscow Oblast (Russia), as well as haplotype H9 of the cultivated plant (Sc1), though this group received poor support in the MP and ML analyses (BP 63, 64%) and strong support only in BI analysis (PP 0.99). The overall topology of MP and BI trees (Fig. 5B) constructed with dataset including thirteen accessions of the I. subser. Sibiricae species retrieved from GenBank was largely similar to those of the trees described above (Fig. 5A). Ten of the thirteen additional accessions of I. sanguinea, I. sibirica, and I. typhifolia were placed together with all specimens of I. subser. Sibiricae in a monophyletic group (BP 100%, PP 1.0). However, the phylogenetic relationships within this clade were unresolved. Only one of three I. sibirica accessions (voucher Mosulishvili G99-12, RSA; see Wilson, 2009) and two (isolates ISD1 and ISD2, Lee & Park, 2013) of six accessions of I. sanguinea from the Republic of Korea were placed outside of the I. subser. Sibiricae clade but clustered with the I. ser. Laevigatae species (Fig. 5B). The sequence divergence (K_S) calculated for two cpDNA regions between Korean accessions of I. sanguinea placed in the I. ser Laevigatae clade and I. sanguinea accessions placed in the I. subser Sibiricae clade was 0.009510 that was comparable with divergence between species in other series of I. sect Limniris (0.00451–0.01223; Boltenkov et al., 2018).

Taxonomic treatment

In the present study we confirm that I. subser. Sibiricae includes only a single variable species, I. sibirica. It is the most widespread Iris species, occurring from Central and Eastern Europe, including northeast Turkey, northern Kazakhstan, and Caucasus, to Siberia, East Asia (northern Mongolia, northern and eastern China, Korean Peninsula, and Japan), and the southern Russian Far East. It is found growing wild in moist meadows along river valleys. It is cultivated worldwide and sometimes naturalized. Morphologically, I. sibirica is distinct from I. subser. Chrysographes species by having shorter bracts (2–6 cm long), a much shorter perianth tube (no more than 0.5 cm long), and green basal leaves. The synonymic list of taxa specified in the present work, including types, is provided below.

Iris sibirica L., Sp. Pl. 1: 39. 1753. ≡ Iris pratensis Lam., Fl. Franç. 3: 498. 1779, nom. illeg. (Art. 52.1, Turland et al., 2018). ≡Biris sibirica (L.) Medik., Staatswirthschaftl. Vorles. Churpfälz. Phys.-Ökon. Ges. Heidelberg, 1: 257. 1791. ≡ Iris stricta Moench, Methodus, 2: 528. 1794, nom. illeg. (Art. 52.1). ≡ Iris angustifolia Salisb., Prodr. Stirp. Chap. Allerton: 44. 1796, nom. illeg. (Art. 52.1). ≡ Xiphion sibiricum (L.) Schrank, Flora 7(2, Beil.): 19. 1824. ≡ Xiphion pratense Parl., Nuov. Gen. Sp. Monocot.: 45. 1854. ≡ Limniris sibirica (L.) Fuss, Fl. Transsilv.: 637. 1866. ≡ Xyridion sibiricum (L.) Klatt, Bot. Zeitung (Berlin), 30: 500. 1872. – Limnirion sibiricum (L.) Opiz, Seznam: 5. 1852, nom. inval. (Art. 38.1). – Iris sibirica var. typica Maxim., Bull. Acad. Imp. Sci. Saint-Pétersbourg, 26: 519. 1880, nom. inval . (Art. 24.3). Type: [Specimen from a cultivated plant]. sibirica 9, HU [Horto Upsaliensis], Herb. Linnaeus (lectotype: designated by Altinordu & Crespo, 2016: 297, LINN! [LINN No. 61.20]).

= Iris orientalis Thunb., Trans. Linn. Soc. London, 2: 328. 1794, nom. illeg. (non Mill., Gard. Dict., ed. 8: Iris No. 9. 1768; Art. 53.1), syn. nov. ≡ Xiphion orientale Schrank, Flora 7(2, Beil.): 19. 1824. ≡ Iris sibirica var. orientalis (Schrank) Baker, J. Linn. Soc., Bot. 16: 139. 1877. ≡ I. extremorientalis Koidz., Bot. Mag. (Tokyo), 40: 330. 1926, nom. nov. (Art. 6.11). Type: Japan. [Note on the upper side]: Iris sibirica, Fl. jap. p. 33, Barin; [Note on the reverse side]: e Japonia, Thunberg s.n., Herb. Thunberg (lectotype: UPS [UPS-THUNB 1144, image!], designated here by E.V. Boltenkov).

= Iris sanguinea Hornem., Hort. Bot. Hafn. 1: 58. 1813, syn. nov. ≡ I. sibirica var. sanguinea (Hornem.) Ker Gawl., Bot. Mag. 39: t. 1604. 1814. ≡ Limniris sanguinea (Hornem.) Rodion., Bot. Zhurn. (Moscow & Leningrad), 92: 551. 2007. – Iris sanguinea Donn, Hort. Cantabrig., ed. 6: 17. 1811, nom. inval. (Art. 38.1) – I. sanguinea var. typica Makino, J. Jap. Bot. 6: 32. 1930, nom. inval. (Art. 24.3). Type: [Specimen from a cultivated plant]. [Handwriting 1]: Iris sanguinea, ex hort. bot. Hafn.; [Handwriting 2]: [Iris sanguinea ] Don., ad I. sibir [ica ]. L. ref. spr., Herb. Hornemann (lectotype: designated by Boltenkov, 2018: 178, C [C10022296, image!]).

= Iris sibirica var. haematophylla Besser, Flora, 17(1, Beibl.): 25. 1834, syn. nov. Type: [Specimen from a cultivated plant]. Iris (sibirica) haematophylla, Dahuria, Fischer s.n., Herb. Lindley (neotype: CGE! [CGE14724 ], designated here by E.V. Boltenkov).

= Iris typhifolia Kitag., Bot. Mag. (Tokyo), 48: 94. 1934, syn. nov. ≡ Limniris typhifolia (Kitag.) Rodion., Bot. Zhurn. (Moscow & Leningrad), 92: 551. 2007. Type: China. [Liaoning Province], Iris sibirica? …14 Aug. 3 [1928], K. Yamatsuta 60 (holotype: TI [image!]).