Introduction

Alisma L. is a cosmopolitan genus of aquatic and wetland plants belonging to the family Alismataceae that is at present thought to contain ten species (Fig. 1). In his monograph of the genus, Samuelsson (1931) recognised six species, namely A. canaliculatum A. Braun et Bouché, A. gramineum Lej., A. lanceolatum With., A. plantago-aquatica L., A. rariflorum Sam. and A. subcordatum Raf. He also accepted several subspecies, including A. plantago-aquatica subsp. orientale (Sam.) Sam., which was originally described as A. plantago-aquatica var. orientale Sam. from Yunnan Province in China and later elevated to species rank by Juzepczuk (1934); Björkqvist (1967, 1968) recognised three species, A. oriantale (Sam.) Juz., A. triviale Pursh A. wahlenbergii (Holmb.) Juz in addition to the above six species. Wang et al. (2010) accepted five of the above-mentioned species (A. canaliculatum, A. plantago-aquatica, A. lanceolatum, A. gramineum plus A. orientale (Sam.) Juz.) plus one additional newly described species, A. nanum D.F.Cui, which is endemic to Xinjiang Province in China. Tanaka (2015) and Kadono (2020), meanwhile, accepted A. canaliculatum, either A. plantago-aquatica var. orientale or A. plantago-aquatica subsp. orientale and A. rariflorum, as well as two recently described varieties from Japan: A. canaliculatum var. azuminoense Kadono & Hamashima and A. canaliculatum var. harimense Makino.

Fig. 1
figure 1

Images of Alisma taxa: a. A. canaliculatum var. azuminoense (Japan); b. A. gramineum (Hungary); c. A. lanceolatum (Malta; TNS:Ito YI1501); d. A. orientale (China; TNS:TNS1177975); e. A. plantago-aquatica (Russia); f. A. rariflorum (Japan)

Full size image

Jacobson and Hedrèn (2007) were the first to assess inter-specific relationships among Alisma species, including A. canaliculatum, A. gramineum, A. lanceolatum, A. orientale, A. plantago-aquatica, A. rariflorum, A. subcordatum, A. triviale and A. wahlenbergii using nuclear DNA (hereinafter nDNA) ITS (hereinafter nrITS), plastid DNA (hereinafter ptDNA) trnL, and RAPD markers. Their ITS tree discerned and identified clades pertaining to A. gramineum A. lanceolatum, A. rariflorum and an A. plantago-aquatica complex including A. orientale (‘plantago-aquatica group’) (Jacobson and Hedrèn 2007). The three species of ‘polyploids’, namely, A. canaliculatum, A. lanceolatum and A. rariflorum, were resolved as non-monophyletic. A tree based on their RAPD data recovered a similar topology, in which the ‘plantago-aquatica group’ and ‘gramineum group’ were both found to be monophyletic, receiving 60% and 100% bootstrap (BS) support, respectively. Meanwhile, the ‘polyploids’ were divided into three lineages, including a lineage in which A. canaliculatum and A. rariflorum are grouped together (85% BS). In their family-level phylogenetic analyses, Chen et al. (2012) shed further light on inter-specific relationships among A. canaliculatum, A. gramineum, A. nanum and A. plantago-aquatica based on nrITS alone, a concatenated ptDNA (matK, psbA and rbcL) tree and a concatenated nrITS and ptDNA tree. However, considerable topological inconsistencies were revealed by Chen et al. (2012), with A. gramineum being resolved as sister to the other three species in their nrITS tree [< 50% maximum likelihood (ML) BS], but A. plantago-aquatica and A. nanum being sister to one other [0.83 posterior probabilities (PP)] in their nrITS tree while A. gramineum being sister to A. nanum (76% ML BS) and A. canaliculatum and A. plantago-aquatica are grouped (86% ML BS) in the ptDNA tree.

Alisma is known to be a polyploid genus (Björkqvist 1968; Jacobson and Hedrèn 2007), with A. lanceolatum documented as 2n = 4x = 26, 28, A. rariflorum as 2n = 4x = 26 and A. canaliculatum as 2n = 4x = 28 (Probatova and Sokolovskaya 1986), 2n = 6x = 40 (Uchiyama 1989) or 2n = 6x = 42 (Wang et al. 1987). The evolutionary path for the generation of polyploids, which may play an important role in the evolution of the genus, is still obscure as neither nrITS nor RAPD trees indicate the origins of polyploidy (Jacobson and Hedrèn 2007).

Alisma canaliculatum is an East Asian species with three varieties: A. canaliculatum var. azuminoense, A. canaliculatum var. canaliculatum (hereinafter A. canaliculatum) and A. canaliculatum var. harimense. The monophyly of the species is yet to be confirmed and infra-specific relationships among the putative subspecies is yet to be investigated except by Jacobson and Hedrèn’s (2007) RAPD tree, in which only a single sample of A. canaliculatum and A. canaliculatum var. harimense were grouped together.

Alisma plantago-aquatica is distributed ubiquitously, occurring in Europe, Russia, East and Southeast Asia, most of North America as well as Australia, New Zealand, Southern Africa and a part of southern South America, which may be introduced (Aston 1973; Cook 2004; Crow and Hellquist 2000; Haynes and Hellquist 2000; Wang et al. 2010). Since Samuelsson (1931), a question remained unresolved about whether to accept an apparent Asian sibling species, A. orientalis, and if so, at which taxonomic rank, i.e., A. plantago-aquatica var. orientale or A. plantago-aquatica subsp. orientale.

The aim of this study was to provide deeper understanding of the evolutionary pathway of the widely distributed polyploid genus Alisma. To do so, we collected six species and two varieties with a primary focus on intra-specific taxa of the A. canaliculatum and A. plantago-aquatica complex and analysed molecular data from multiple loci, including two nDNA markers and ptDNA. Discussion is made in regard to polyploidy and species delimitation based on the resulting molecular phylogenetic trees.

Materials and methods

Taxon sampling

Samples of Alisma were collected in the field or obtained from herbarium specimens with a special focus on Asia, an apparent diversity hotspot of the genus (Fig. 2; Table 1). For specimen identification, we used the taxonomic treatments of Samuelsson (1931), with cross referencing to the following local floras: Wang et al. (2010), Tanaka (2015) and Kadono (2020). An Asian taxon occasionally recognised as either A. plantago-aquatica var. orientale or A. plantago-aquatica subsp. orientale (e.g., Kadono 2020; Tanaka 2015) was tentatively treated as A. orientale following Wang et al. (2010). In total, six species and two varieties of Alisma were collected: A. canaliculatum var. azuminoense (two samples); A. canaliculatum var. canaliculatum (eight samples); A. canaliculatum var. harimense (two samples); A. gramineum (two samples); A. lanceolatum (two samples); A. orientale (three samples); A. plantago-aquatica (six samples); A. rariflorum (two samples) (Table 1). Two North American species, A. triviale and A. subcordatum, occasionally treated as varieties of a widespread A. plantago-aquatica (Haynes and Hellquist 2000), a European species, A. wahlenbergii, and A. nanum, a rare and enigmatic species endemic to Xinjiang, China (Wang et al. 2010), were not included. Baldellia Parl., Damasonium Mil., Luronium Raf. and Sagittaria L., representatives of closely related genera were chosen as outgroup taxa following Chen et al. (2012) and Ross et al. (2016). Sequences of D. alisma and L. natans used in Les et al. (1997) and Les and Tippery (2013) were downloaded from GenBank and included in subsequent phylogenetic analyses.

Fig. 2
figure 2

Locality of Alisma samples collected and used in this study. Alisma canaliculatum (Purple closed circles); A. canaliculatum var. azuminoense (Orange open circle); A. canaliculatum var. harimense (Blue open double circles); A. gramineum (Pale blue triangles); A. lanceolatum (Green squares); A. orientale (Pink open stars); A. plantago-aquatica (Red closed stars); A. rariflorum (Forest inverted triangles). Note that one locality of A. plantago-aquatica in Japan is not shown (Garden origin). The symbols for each taxa are shown in the following molecular phylogenies (Figs. 3–7)

Full size image
Table 1 Specimen and voucher information for the taxa of Alisma included in this study
Full size table

DNA extraction, amplification and sequencing

Total genomic DNA was extracted from silica gel-dried leaf tissue using the CTAB method described by Ito et al. (2010). We chose two nDNA markers, i.e., the multi-copy nrITS and the low-copy nuclear-encoded phyA of the phytochrome gene family, and four regions of ptDNA (matK, ndhF, psbA-trnH and rbcL). The six regions were PCR amplified with the following primers: ITS-4 and ITS-5 for nrITS (Baldwin 1992); Alis-phyA-F (5′-ATCCTTGCCTGGTGGGAGTA) and Alis-phyA-R (5′-GAGGATCTAGGATGCATTCTG), with a simultaneous use of Alis-phyA-2F (5′-GCAAGAGCAACCAAAGAGGA) and Alis-phyA-3F (5′-GCACAAGTCTTTGCTTTACA) for phyA; trnK-Hy2F and trnK-Bly3R (Ito et al. 2019) with a simultaneous use of Alis_matK-2F (5′-TCCGAGCCAAAGTTCTAGCGC), Alis_matK-2R (5′-GATTAGTGCCTAATCCGGGA), Alis_matK-3F (5’-GCGAGAATTGATTTTCCTTG), and Alis_matK-3R (5′-CCAGTTTCATACTTGTGAAACG) for matK; ndhF-F2 (Oxelman et al. 1999) and ndhF-1955R.re (Ito et al. 2017) modified from the primer ndhF-1955R published by Olmstead and Sweere (1994) for ndhF; psbAF and trnHR (Sang et al. 1997) for psbA-trnH, rbcL-F1F (Wolf et al. 1994) and rbcL-1379R (Little and Barrington 2003) for rbcL. PCR amplification was conducted using TaKaRa Ex Taq polymerase (TaKaRa Bio, Shiga, Japan). PCR cycling conditions were 94 °C for 60 s; then 30 cycles of 94 °C for 45 s, 52 °C for 30 s, 72 °C for 60 s, with a final extension of 72 °C for 5 min. PCR products were cleaned using ExoSAP-IT purification (GE Healthcare, Piscataway, NJ, USA) and amplified using ABI PRISM Big Dye Terminator ver. 3.1 (Applied Biosystems, Foster City, CA, USA) with the same primers as those used for the PCR amplifications. Cloning and subsequent sequencing of clones followed Ito et al. (2010); i.e., for PCR products, in which overlapping double peaks were found at the same sites for complementary strands in the electropherograms, we cloned using a TOPO TA Cloning kit for Sequencing (Invitrogen, Carlsbad, California, USA) at least 16 clones per sample and their sequences were determined using the same procedure as that used in the first PCR followed by direct sequencing. For the cloned sequences, nucleotides that were not detected by direct sequencing were regarded as PCR errors. Chimeric clones were examined by eye and then eliminated. DNA sequencing was performed with an ABI PRISM 3500XL Genetic Analyzer (Applied Biosystems). Automatic base-calling was checked by eye using Genetyx-Win ver. 3 (Software Development Co., Tokyo, Japan). The sequences generated and their metadata were submitted to the DNA Data Bank of Japan (DDBJ), which is a GenBank data provider (Table 1).

Data analysis and molecular phylogenetic analyses

We assembled three datasets of Alisma: nrITS; phyA; ptDNA (matK, ndhF, psbA-trnH and rbcL). All 27 samples were included in the nrITS and ptDNA datasets while 19 of them were included in the phyA dataset (Table 1). Sequences were aligned using Mafft ver. 7.058 (Katoh and Standley 2013) and then inspected manually.

Phylogenetic inference was performed using maximum parsimony (MP), Bayesian inference (BI; Yang and Rannala 1997) and maximum likelihood (ML). A heuristic MP search was performed in PAUP* ver. 4.0b10 (Swofford 2002) with 100 random addition sequence replicates and tree bisection-reconnection (TBR) branch swapping, with the MulTrees option in effect. The MaxTrees option was set at 100,000. MP bootstrap analyses (Felsenstein 1985) were performed using 1,000 replicates with TBR branch swapping and a random addition sequence. The MaxTrees option was set at 1,000 to avoid local optima.

Bayesian analyses were conducted using MrBayes ver. 3.2.6 (Ronquist et al. 2012), after evaluating the best model in MrModeltest ver. 3.7 (Nylander 2002), which were GTR + G for plastid DNA, GTR + G for nrITS and HKY + G for phyA. Analyses were run for three million generations, sampling every 100th generation and discarding the first 25% as burn-in. Convergence and effective sampling sizes (ESS) of all parameters were checked in Tracer ver. 1.6 (Rambaut et al. 2014).

For the ML analysis, we used the RAxML BlackBox online server (https://raxml-ng.vital-it.ch/#/), which supports GTR-based models of nucleotide substitution (Stamatakis 2006). The ML search option was used to find the best-scoring tree after bootstrapping. Statistical support for branches was calculated by rapid bootstrap analyses of 100 replicates (Stamatakis et al. 2008).

Nodes were recognised as strongly, moderately or weakly supported based on the following thresholds: 100% MP BS, 100% ML BS, 1.0 posterior probability (PP); ≥ 90% MP BS, ≥ 90% ML BS, ≥ 0.95 PP; and ≥ 70% MP BS, ≥ 70% ML BS, ≥ 0.90 PP, respectively.

Analysis with Homologizer

In order to phase the alleles of the two nDNA markers (nrITS and phyA) and ptDNA, we used Homologizer (Freyman et al. 2022) implemented in RevBayes (Höhna et al. 2016). Homologizer inferred multi-locus phylogeny while assigning of each gene copy, for each locus, into each of the available subgenomes. The potential parental taxa with homozygous nDNA genotypes were fixed firstly to infer the pattern of the subgenome (or allele) evolution of this group. Each locus (dataset) was modelled with an independent GTR substitution model, with Dirichlet priors on the exchangeability parameters and stationary frequencies, a uniform prior on topology, and exponential priors (mean = 0.1) on branch lengths.

The MCMC was run four times independently; each for 10,000 generations and the first 25% of trees were discarded as burn-in. Convergence was checked by Tracer ver. 1.6 (Rambaut et al. 2014) to see if the ESS was over 200. The obtained maximum a posteriori (MAP) tree was then rooted by FigTree ver. 1.3.1 (Rambaut 2009) and used for the subsequent analysis.

The phasing estimates from a Homologizer MCMC analysis was summarised and plotted using the R (R Core Team 2021) package RevGadgets (Tribble et al. 2022). The first 10% of samples were discarded prior to summarising the posterior distribution.

Multi-locus species delimitation analysis

A Bayesian coalescent method of species delimitation was performed using STACEY (species tree estimation using DNA sequences from multiple loci; Jones 2017), which is an extension of DISSECT (Jones et al. 2015). STACEY was implemented in BEAST ver. 2.4.4 (Bouckaert et al. 2014; Drummond and Rambaut 2007; Drummond et al. 2006). We ran STACEY using a multilocus data set (nrITS, phyA and ptDNA) with all ingroup taxa; outgroup species were excluded to avoid rate differences and hidden substitutions between ingroup and outgroup species (B. Oxelman, personal communication, November 22, 2016). Since some accessions possess multiple nrITS and phyA sequences, we used Homologizer results to see which heterogeneous nrITS sequence should correspond to those in phyA with the highest posterior probabilities. We performed two independent runs of ten million generations of the MCMC chains, sampling every 1,000 generations. Convergence of the stationary distribution was checked by visual inspection of plotted posterior estimates using Tracer ver. 1.6 (Rambaut et al. 2014). After discarding the first 1,000 trees as burn-in, the samples were summarised in the maximum clade credibility tree using TreeAnnotator ver. 1.6.1 (Drummond and Rambaut 2007) with a posterior probability limit of 0.5 and summarising of mean node heights. The results were visualised using FigTree ver. 1.3.1 (Rambaut 2009).

The alignments, supplementary tables, xml files, scripts and resulting log files, intermediate files and trees were deposited in Mendeley Data (https://doi.org/10.17632/rsgx77r2rm.1).

Results

NrITS phylogeny

The nrITS dataset of Alisma included 769 aligned characters, of which 80 were parsimony informative. Analysis of this dataset yielded the imposed limit of 100,000 MP trees (tree length = 278 steps; consistency index = 0.89; retention index = 0.92). The strict-consensus MP tree, the RAxML tree, and the MrBayes BI 50% consensus tree showed no incongruent phylogenetic relationships; thus only the MrBayes tree is presented here (Fig. 3).

Fig. 3
figure 3

MrBayes tree of Alisma based on nuclear ITS (nrITS). Branch lengths are proportional to molecular divergence among accessions. Numbers at the branches indicate bootstrap support (BS) calculated in maximum parsimony (MP BS) and maximum likelihood (ML BS) analyses and Bayesian prior probabilities (PP), respectively. Polyploids are shown in bold while diploids and those with no ploidy information are shown in regular. Note that two varieties of A. canaliculatum are shown as A. c. var. azuminoense and A. c. var. harimense, respectively

Full size image

The monophyly of Alisma is supported (98% MP BS, 55% ML BS, 1.0 PP). Alisma lanceolatum (clade I) is sister to all other taxa (81% MP BS, 48% ML BS, 1.0 PP); A. gramineum (clade II) is then sister to the core group (64% MP BS, 44% ML BS, 0.95 PP). The core group is divided into two lineages: one containing A. canaliculatum var. azuminoense (clade V), A. canaliculatum var. harimense (clade VI), A. rariflorum (clade III) and clones of A. canaliculatum_TD3957, A. canaliculatum_TD4158 and A. canaliculatum_TNS01178068 (clade IV) (99% MP BS, 83% ML BS, 1.0 PP), and the other containing A. plantago-aquatica, A. orientale and the other clones of A. canaliculatum_TD3957, A. canaliculatum_TD4158 and A. canaliculatum_TNS01178068 as well as the rest of A. canaliculatum (100% MP BS, 92% ML BS, 1.0 PP). The former lineage is sub-divided into A. rariflorum (clade III) plus all the remaining accessions (clades IV-VI; 95% MP BS, 83% ML BS, 1.0 PP). In the latter lineage, three accessions of A. orientale (clade VII) and seven accessions of A. canaliculatum (clade VIII) are each supported as monophyletic (85% MP BS, 85% ML BS, 1.0 PP and 64% MP BS, 67% ML BS, 0.99 PP, respectively).

PhyA phylogeny

The phyA dataset of Alisma included 921 aligned characters, of which 29 were parsimony informative. Analysis of this dataset yielded the imposed limit of 100,000 MP trees (tree length = 158 steps; consistency index = 0.94; retention index = 0.93). The strict-consensus MP tree, the RAxML tree, and the MrBayes BI 50% consensus tree showed no incongruent phylogenetic relationships. Thus, only the MrBayes tree is presented here (Fig. 4).

Fig. 4
figure 4

MrBayes tree of Alisma based on nuclear phytochrome A gene (phyA). Branch lengths are proportional to molecular divergence among accessions. Numbers at the branches indicate bootstrap support (BS) calculated in maximum parsimony (MP BS) and maximum likelihood (ML BS) analyses and Bayesian prior probabilities (PP), respectively. Asterisks represent branches not recovered in maximum parsimony or maximum likelihood analyses. Polyploids are shown in bold while diploids and those with no ploidy information are shown in regular. Note that two varieties of A. canaliculatum are shown as A. c. var. azuminoense and A. c. var. harimense, respectively

Full size image

Alisma lanceolatum (Singleton I) is resolved as sister to all other accessions in the RAxML tree (52% ML BS; data not shown) or weakly clustered with clones of A. canaliculatum, A. canaliculatum var. azuminoense, A. canaliculatum var. harimense and A. rariflorum in the MrBayes tree (0.56 PP). Some well-supported clades are retrieved: a clade containing A. plantago-aquatica and one of three clones of A. canaliculatum_TNS01178068 (57% MP BS, 43% ML BS, 0.54 PP), in which two accessions of A. plantago-aquatica (clade IX) and two accessions of A. orientale plus one of the three clones of A. canaliculatum_TNS01178068 (clade VII) are clustered (66% MP BS, 82% ML BS, 0.87 PP and 64% MP BS, 57% ML BS, 0.87 PP, respectively); a clade containing one of two clones of A. canaliculatum and A. canaliculatum var. azuminoense and another clone of A. canaliculatum_TNS01178068 (clade IV; 0.51 PP); a clade containing second copies of A. canaliculatum and the last clone of A. canaliculatum_TNS01178068 (clade VIII; 76% MP BS, 69% ML BS, 0.76 PP) which is further clustered with second clones of A. rariflorum (clade III) (81% MP BS, 76% ML BS, 1.0 PP).

PtDNA phylogeny

The ptDNA dataset of Alisma comprising four genes included 4,830 aligned characters, of which 43 were parsimony informative. Analysis of this dataset yielded the imposed limit of 100,000 MP trees (tree length = 344 steps; consistency index = 0.98; retention index = 0.96). The strict-consensus MP tree, the RAxML tree, and the MrBayes BI 50% consensus tree showed no incongruent phylogenetic relationships; thus, only the MrBayes tree is presented here (Fig. 5).

Fig. 5
figure 5

MrBayes tree of Alisma based on plastid DNA (matK, ndhF, psbA-trnH and rbcL). Branch lengths are proportional to molecular divergence among accessions. Numbers at the branches indicate bootstrap support (BS) calculated in maximum parsimony (MP BS) and maximum likelihood (ML BS) analyses and Bayesian prior probabilities (PP), respectively. Polyploids are shown in bold while diploids and those with no ploidy information are shown in regular. Note that two varieties of A. canaliculatum are shown as A. c. var. azuminoense and A. c. var. harimense, respectively

Full size image

The monophyly of Alisma is supported (100% MP BS, 96% ML BS, 1.0 PP). Alisma is broadly divided into two lineages: one containing A. gramineum (clade II) and the A. plantago-aquatica complex including A. orientale (59% MP BS, 92% ML BS, 0.92 PP) and the other containing A. canaliculatum, A. canaliculatum var. azuminoense, A. canaliculatum var. harimense, A. lanceolatum and A. rariflorum (77% MP BS, 91% ML BS, 1.0 PP). Alisma canaliculatum, A. canaliculatum var. azuminoense, and A. canaliculatum var. harimense and A. rariflorum (clade III) are further supported in the latter lineage (100% MP BS, 100% ML BS, 1.0 PP). In this latter lineage, two accessions of A. canaliculatum var. azuminoense are grouped with A. canaliculatum_TD4290 and A. canaliculatum_YI1469 (clade V; 60% MP BS, 70% ML BS, 0.99 PP).

Homologizer analysis

The maximum a posteriori (MAP) phylogeny of Alisma obtained from Homologizer analysis is shown in Fig. 6. With A. lanceolatum as a sister lineage to the remaining species (0.14 PP), two major lineages were recovered: one with A. canaliculatum plus A. rariflorum (1.0 PP) and the other with A. canaliculatum, A. gramineum, A. plantago-aquatica complex including A. orientale (0.51 PP; Fig. 6). The former lineage was sub-divided into A. canaliculatum (1.0 PP) and A. rariflorum (1.0 PP). With A. gramineum as a sister to all remaining samples (0.44 PP), the latter lineage was sub-divided into A. plantago-aquatica complex including A. orientale plus one of three A. canaliculatum_TNS01178068_A (0.23 PP) and a clade containing two species (0.11 PP), which was split into A. canaliculatum (0.32 PP) and A. rariflorum (0.92 PP).

Fig. 6
figure 6

The maximum a posteriori (MAP) phylogeny of Alisma obtained from Homologizer analysis. The phasing of gene copies into subgenomes is summarised on the phylogeny. Colour of circles on each node indicate posterior probabilities. The columns of the heatmap each represent a locus, and the joint MAP phase assignment is shown as text within each box. Each box is coloured by the marginal posterior probability of the phase assignment. In the sample labels, capital A, B, and C, indicate subgenomes, and lowercase letters following the accession numbers indicate haploid “individuals” within the sampled diploids (i.e., those diploids are heterozygous). Copy names with a “BLANK” suffix indicate missing sequences (e.g., a subgenome that was present in some loci but not retrieved for others). Polyploids are shown in bold while diploids and those with no ploidy information are shown in regular. Note that two varieties of A. canaliculatum are shown as A. c. var. azuminoense and A. c. var. harimense, respectively

Full size image

Species delimitation using STACEY

SpeciesDelimitationAnalyser generated 1,571 patterns of clusters from the MCMC runs. Classifications with eight Alisma species or minimal clusters received highest posterior probabilities (0.10 PP). Under the classification, the following three clusters were conspecific: A. canaliculatum var. harimense_TNS766110_A, A. orientale (three samples) plus A. canaliculatum_TNS01178068_A and A. plantago-aquatica (six samples). However, the similarity scores among these clusters were too low to support the recognition of distinct species: the similarity score between A. plantago-aquatica (six samples) and A. canaliculatum (eight samples) plus A. canaliculatum var. azuminoense_TD4643_A was 0.14–0.44, whereas that between A. plantago-aquatica (six samples) and A. orientale (three samples) plus A. canaliculatum_TNS01178068_B was 0.23–0.35. In addition, the similarity score of 0.43–0.81 for all A. plantago-aquatica (six samples) samples was close to that of 0.40–0.89 for all A. canaliculatum samples including A. canariculatum var. harimense (eight samples) (Table S1). Therefore, the above-mentioned three clusters were treated as species or minimum clusters (Fig. 7).

Fig. 7
figure 7

Species or minimal clusters tree (Jones et al. 2015) of Alisma based on a multi-locus data set (nrITS, phyA and ptDNA) with the similarity matrix showing posterior probabilities for pairs of individuals belonging to the same cluster. Thickness of branches denotes posterior probabilities obtained from BEAST 2. The squares in the matrix represent posterior probabilities (white = 0, black = 1) for pairs of individuals that belonged to the same cluster. Note that two varieties of A. canaliculatum are shown as A. c. var. azuminoense and A. c. var. harimense, respectively

Full size image

Discussion

Phylogeny of Alisma revisited

Our nrITS phylogeny showed no major conflicts with previous phylogenetic insights provided by Jacobson and Hedrèn (2007) and Chen et al. (2012), with A. lanceolatum and A. gramineum being confirmed as two early-diverging species. Alisma rariflorum was resolved as sister to the remaining taxa of the core clade, including A. canaliculatum and A. plantago-aquatica, by Jacobson and Hedrèn (2007), which may correspond to the relationship between A. canaliculatum (clade VIII) and A. plantago-aquatica uncovered by our phylogeny (Fig. 3). This suggests that Jacobson and Hedrèn (2007) failed to uncover the true relationships among these taxa owing to the omission of A. canaliculatum var. azuminoense and A. canaliculatum var. harimense (clades V and VI in Fig. 3).

With eight samples' sequences lacking, our phyA phylogeny successfully and clearly distinguished species of Alisma from each other, notably A. orientale and A. plantago-aquatica, though the inter specific phylogenetic relationship was scarcely resolved (Fig. 4). Cloned heterogeneous phyA sequences did not cluster with one another and were instead found to be only distantly related, e.g., clades IV and VIII (Fig. 4).

Chen et al. (2012) presented a ptDNA phylogeny of four species of Alisma using matK, psbA and rbcL, in which A. gramineum was revealed as sister to A. nanum while A. canaliculatum and A. plantago-aquatica are grouped together. As long as their accessions of A. canariculatum, A. gramineum and A. plantago-aquatica showed no or limited differences when compared to ours (data not shown), the topological differences between their and our ptDNA phylogenies may be attributed to their single accession of A. nanum in Chen et al. (2012).

In our MAP phylogeny, based on nrITS, phyA and ptDNA data, Alisma canaliculatum, A. canaliculatum var. azuminoense, A. canaliculatum var. harimense and A. rariflorum are placed twice in the phylogeny, indicating their allopolyploid origin (Fig. 6). Monophyly of each copy of A. canaliculatum including two varieties was supported (Fig. 6). The Alisma planago-aquatica complex including A. orientale is resolved as monophyletic, in which all A. orientale samples are further clustered (Fig. 6). Although A. plantago-aquatica is not resolved as monophyletic in the MAP phylogeny, the six samples are sufficiently differentiated from the three samples of A. orientale in our Stacey analysis (see below).

Species delimitation and geographic implications for Alisma plantago-aquatica complex

Since Linneaus (1753) described Alisma plantago-aquatica as the type species of the genus, this species has proved to be the most ubiquitous. A sibling taxon, A. plantago-aquatica var. orientale, a basionym of A. orientale, was described from Yunnan, China and is occasionally recognised in Asia as a smaller form of A. plantago-aquatica in Europe. For example, Juzepczuk (1934) states, “Flowers ca. 6 mm in diameter, carpels 1.5 – 2 mm long” with regards to A. orientale and, “Flowers ca. 1 cm in diameter, carpels commonly 2–3 mm long” with regards to A. plantago-aquatica. Similarly, Wang et al. (2010) state, “Petals marginally undulate; carpels irregularly arranged; styles ca. 0.5 mm” in relation to A. orientale and, “Petals marginally denticulate; carpels regularly arranged; styles 0.7–1.5 mm” in relation to A. plantago-aquatica (Table 2). The recognised distribution of these entities is segregated into Asia including southern central Asia for A. orientale and Europe including northern central Asia for A. plantago-aquatica (Juzepczuk 1934). However, according to some authors their ranges overlap from India to Mongolia and Japan (Wang et al. 2010).

Table 2 Morphological comparisons of Alisma species used in this study. Important diagnostic characters are shown in bold
Full size table

In the present study, three samples of A. orientale collected from China, Myanmar, and Vietnam are discerned from A. plantago-aquatica from Europe, Japan, Korea and New Zealand (Fig. 7). A flowering specimen of A. orientale_TNS1177975 from China exhibited irregularly arranged carpels and shorter styles. Though sampling was limited, this study further indicates apparent diagnostic characters between the two species in leaf apex, i.e., acuminate in A. orientale and less acuminate in A. plantago-aquatica.

Alisma plantago-aquatica is common in East Asia but not so in Southeast Asia, and it tends to be restricted to higher elevations, e.g., a record from Thailand was documented at alt. 1200 m in Chiang Mai, northern Thailand (Haynes 2001). Three of our A. orientale samples are all from Southeast Asia’s highlands (Southeast Asian Massif): one from the Shan Plateau, Northeast Myanmar, one from the Yunnan–Guizhou Plateau, Southeast China, and the other from a mountainous region in northwest Vietnam. From an evolutionary point of view, A. orientale is apparently endemic to the Southeast Asian Massif and may have originated from peripatric speciation (Futuyma 2005) at the southern edge of the distribution range of the widely-distributed sibling species, A. plantago-aquatica.

Allotetraploid origin of Alisma canaliculatum and A. rariflorum

Alisma canaliculatum in East Asia is known to be a tetraploid or a hexaploid (Jacobson and Hedrèn 2007; Probatova and Sokolovskaya 1986; Uchiyama 1989; Wang et al. 1987). Alisma rariflorum, endemic to Japan, is another species inferred to be a tetraploid by flow cytometry (Jacobson and Hedrèn 2007). The present study included 12 samples of A. canaliculatum from Japan including two samples each of A. canaliculatum var. azuminoense and A. canaliculatum var. harimense. All of these are divided into two monophyletic clades in our phylogeny, with high or moderate support (Fig. 6; but note the third phyA copy for A. canaliculatum_TNS01178068). Similarly, two samples of A. rariflorum from Japan are also divided into two clades, both of which are sister to those of A. canaliculatum (Fig. 6). The results strongly indicate that the two species are of allotetraploid origin, and the diploid progenitors are now extinct, respectively. Given that A. rariflorum is endemic to Japan, A. canaliculatum may have originated in Japan and then expanded its distribution to the Korean peninsula and to mainland China.

Alisma canaliculatum_TNS01178068 from Nagano, Japan, is the only sample that shows three phyA clones, indicating its hexaploid origin (Figs. 4, 6). Its evolutionary origin may be explained by an ancient cross between allotetraploid A. canaliculatum and diploid A. orientale. Guizhou or Sichuan Provinces in China are potential areas of origin of this event, since A. canaliculatum is currently distributed there (Wang et al. 2010) and A. orientale occurs in the neighbouring province, Yunnan.

In Japan, two local infraspecific taxa are recognised under A. canaliculatum, the taxonomic validity and evolutionary relationships of which remain unexplored. Alisma canaliculatum var. harimense was described from Hyogo Prefecture by Makino (1940) and is distinguished by its narrower leaves, i.e., 3–10 mm in width. Alisma canaliculatum var. azuminoense was described from Nagano Prefecture by Kadono and Hamashima (1988) based on its densely clustered inflorescences. The present study revealed their unique phylogenetic position in the nrITS phylogeny (clades V and VI in Fig. 3), yet the other analyses showed similarities with other A. canaliculatum samples (Figs. 4–6). Considering their unique morphology, we follow the current infraspecific taxonomy of A. canaliculatum in recognising all three varieties.

Eight samples of A. canaliculatum var. canaliculatum are not resolved as monophyletic (Fig. 6). Alisma canaliculatum_YI1469 and A. canaliculatum_TD4290 from Nagano and Toyama prefectures, Central Japan, are distinguished from the other samples from Fukushima, Ibaraki and Tochigi prefectures, Northern Japan, in the ptDNA phylogeny (Fig. 5). Given that the two large areas are geographically disjunct and the haplotype is identical to that of A. canaliculatum var. azuminoense from Nagano prefecture (Fig. 5), it may be possible that the diploid progenitor inhabited the mountainous area in Central Japan including the Hokuriku region.

Notes on Alisma lanceolatum

Alisma lanceolatum is known as a tetraploid by chromosome observations (e.g., Uchiyama 1989) and by flow cytometry (Jacobson and Hedrèn 2007). While the other polyploids of Alisma possess heterogeneous phyA sequences, no such heterogeneity was recovered in the two samples of A. lanceolatum (Figs. 3, 4). Future molecular phylogenetic study using samples of A. lanceolatum for which chromosome numbers have been verified by chromosome counts may yet reveal further details of this species’ evolution.

Taxonomic treatment

Alisma orientale (Sam.) Juz. In Komarov, Fl. URSS. 1: 281. (1934).

Type. China. North Yunnan. Alt. ca. 1950 m, 8 Apr. 1922, H. Smith 1587 (Holotype: UPS).

Diagnosis. Irregular carpel arrangement, marginally undulate petals and styles ca. 0.5 mm.

Distribution. China (Yunnan), Myanmar (Shan), Vietnam (Lao Cai).

Note. As is mentioned in Plants of Central Asia (Grubov et al. 2002), “Typical specimens of A. orientale are mainly confined to the south-eastern part of the distribution,” the species’ distribution may be southeast Asia. A record under the name of A. plantago-aquatica at alt. 1200 m in Chiang Mai, northern Thailand (Haynes 2001) may belong to this species. Specimen records at Zhou Cheng, Yunnan (alt. 2010 m), China and Ha Giang (alt. 1310 m) and Lai Chau (alt. 1000–1100 m), Vietnam may also belong to this species. Illustration of Manchurian Water-plants (Sato 1942) depicts irregularly arranged carpels and marginally undulate petals, suggesting the occurrence of A. orientale also in Dalian, north-eastern China.

Conclusions

With an aim to infer its evolutionary history and species delimitation, we analysed multi-locus molecular data from six putative species and two varieties of Alisma in the world. Apparent polyploids possess each distinct heterogeneous genome, strongly indicating their allopolyploid origin. A ubiquitous, widespread species and its sibling, which has occasionally been treated as a variety or subspecies of the former, are both treated at species rank based on our species delimitation analysis. The latter species is apparently endemic to the Southeast Asian Massif and may have been originated as a case of peripatric speciation. A future study will focus on the geographic distribution of A. canaliculatum var. canaliculatum in mainland China as the variety is divided into two types in Japan, each of which are geographically and ecologically slightly differentiated, probably adapted to highlands and lowlands in the mainland of Japan.