Introduction

Rheophytes are confined to streambeds subject to flash floods; they grow up to flood levels, but not beyond the reach of regularly occurring flash floods (van Steennis 1981, 1987). Rheophytes have adaptive morphology allowing them to grow even when frequently submerged in strong swift-moving water. One such adaptation is stenophylly (narrowness of leaves or leaflets), which reduces the pressure of violent water currents (van Steennis 1981; Imaichi and Kato 1992). Osmunda lancea Thunb. is an obligate rheophyte endemic to Japan (Ohwi 1957; Tagawa 1959). The fern genus Osmunda subgenus Osmunda comprises O. lancea and two other species, O. japonica Thunb. and O. regalis L. (Yatabe et al. 2005). Molecular phylogenetic trees constructed based on chloroplast DNA sequence data suggest that O. lancea is the species most closely related to O. japonica (Yatabe et al. 1999; Metzgar et al. 2008). Osmunda japonica is distributed widely throughout the dryland of East Asia, and O. regalis is cosmopolitan (Kato 2007). It is very likely that O. lancea was derived from O. japonica and adapted to novel environments.

Osmunda × intermedia (Honda) Sugimoto, a natural hybrid between O. japonica and O. lancea, occurs in various rivers across Japan (Shimura 1972; Shimura and Matsumoto 1977). The hybrid shows morphological intermediacy between the parental species in its leaf shape. It usually grows most abundantly in the upper rheophytic zone and frequently exceeds the parent species in vegetative vigor. Zonation is frequently observed in the distribution of O. lancea, O. × intermedia and O. japonica along the riverbank. Therefore, the leaf morphology is considered to be subjected to strong selection pressure. It is, however, unclear if the parental species are genetically distinct, or if the wild hybrid populations consist only of primary F1 hybrids because O. × intermedia is semi-fertile and has moderate spore germination rates (4–11%: Shimura 1964). The present study examined the genetic structure of populations consisting of O. × intermedia and the two parental species by analyzing chloroplast DNA sequences and ten nuclear DNA markers including the partial sequence of a glyceraldehyde-3-phosphate dehydrogenase gene (GapCp), six cleaved amplified polymorphic sequence (CAPS) markers and three simple sequence repeats (SSRs) markers, which were developed from an expressed sequence tag (EST) library.

Materials and methods

Plant materials

Living plant materials of Osmunda lancea, O. japonica and O. × intermedia were collected from the localities listed in Table 1. In a Hanno population (Population H) in Saitama Prefecture, comprising the two parental species and their putative hybrids, all of the 47 sporophytes growing together were collected; their distribution is shown in Fig. 1. Similar populations may occur in the neighborhood of Population H. In addition to the 47 sporophytes, 8 sporophytes of O. lancea were also collected from the adjacent sites (less than 10 m away from the population of Fig. 1) along the same ravine and treated as members of the same population. Another 20 and 24 sporophytes were also collected from a population at the Shojin River in Shizuoka Prefecture (Population S), which is about 100 km distant from Population H; and a population from the Kyushu area of Miyazaki Prefecture (Population K), which is located on another island no less than 700 km distant from Population H and Population S. Vouchers specimens for all the sporophytes of the three populations (Population H, S and K), and an additional nine individuals from other localities, have been deposited in the National Museum of Nature and Science Herbarium (TNS; Table 1).

Table 1 Species, locality and voucher information of plant materials used for rbcL sequencing and their GenBank database accession numbers
Fig. 1
figure 1

Distribution of Osmunda japonica (open circles), Osmunda × intermedia (grey circles) and Osmunda lancea (solid circles) in Hanno mixed population. Numbers in parentheses are the numbers of individuals in subpopulations

For each sporophyte of Population H, and one very young sporophyte of O. lancea, five pinnules were picked out randomly in order to measure morphological characters as indicators of stenophylly. The number of ultimate veinlets of the most basal basiscopic vein was counted, and the angle of pinnule-base was measured (Fig. 2).

Fig. 2
figure 2

Morphological characters of pinnules measured in this study. a O. japonica, b O. lancea. Arrows Ultimate veinlets of the most basal basiscopic vein. Diagrams on the right indicate measured angles. Bars 1 cm

DNA was extracted from all sporophytes using a QIAGEN DNeasy mini kit (QIAGEN, Valencia, CA) after drying the fresh material with silica gel.

Sequencing of chloroplast DNA

The atpB-rbcL spacer, rbcL, rbcL-accD spacer, accD and trnL-F intergenic spacer were analyzed using the primers listed in Table 2. PCR was performed using a thermal cycler (Perkin-Elmer 9700, Applied Biosystems, Foster, CA) with Ex Taq DNA polymerase (Takara, Tokyo, Japan). Amp direct (Shimadzu, Kyoto, Japan) was used for efficient amplification, substituting the Ex buffer. PCR products were purified with ExoSAP-IT (USB Corporation, Cleveland, OH) following the instruction manual. The cycle sequencing samples were purified by ethanol precipitation. Sequencing was conducted using an ABI PRISM 3130xl Genetic Analyzer (Applied Biosystems). The obtained sequences were assembled using SeqMan II (Dnastar, Madison, WI).

Table 2 Description of the molecular markers used in this study

Sequencing of GapCp

Total RNA was extracted from fresh leaves of Osmunda japonica and O. regalis and Osmundastrum cinnamomeum cultivated in Tsukuba Botanical Garden with PureLink Plant RNA Reagent (Invitrogen, Carlsbad, CA) and purified using a QIAGEN RNeasy Mini kit (QIAGEN). Single-stranded cDNA was synthesized using the 3′ RACE System for Rapid Amplification of cDNA Ends (Invitrogen). To amplify GapCp genes, a primer, GAAKA (5′-GGNGCNGCNAARGCNCTNGC-3′), was designed based on registered sequences of GapCp in the DDBJ/EMBL/NCBI database. The first round of PCR and the nested PCR were conducted using GAAKA and the universal amplification primer (UAP) provided in the 3′ RACE System, WYDNE (Petersen et al. 2003) and UAP, respectively. The PCR products were cloned into a pGEM-T Vector (Promega, Madison, WI) and sequenced. Some specific signatures of amino acid sequences of glyceraldehyde-3-phosphate dehydrogenase shown by Petersen et al. (2003) were observed in the sequences obtained, suggesting that the sequences were identical to nuclear GapCp. Based on these sequences, a primer set, OgapF1 (5′-GGTGCTGCCAGAATTGAATGGGAA-′3) and OgapR1 (5′-CGAAACCGTTGTTCAAAGCAATACCTG-3′), was designed. Partial GapCp fragments were then amplified, cloned, and sequenced using the primer set using genomic DNA of the same samples of O. japonica and O. regalis extracted using a QIAGEN DNeasy Plant Mini Kit (QIAGEN). Based on the sequences, a primer set, listed in Table 2, was designed in order to amplify fragments without insertion or deletion among the samples of O. lancea, O. × intermedia and O. japonica. Sequencing was conducted as described for sequencing of chloroplast DNA.

cDNA library and cDNA sequencing

From the total RNA extracted from the individual of Osmunda lancea cultivated in Tsukuba Botanical Garden, cDNA was synthesized using CDS-3M adapter (Evrogen, Moscow), SMART IV Oligonucleotide (Takara) and PrimeScript Reverse Transcriptase (Takara) according to the instruction manual of the Trimmer-Direct cDNA normalization kit (Evrogen). After amplification of cDNA using the 5′ PCR primer included in the SMART cDNA library construction kit (Takara) and Advantage 2 polymerase (Takara), cDNA was normalized using the Kamchatka crab duplex-specific nuclease included in the Trimmer-Direct cDNA normalization kit (Evrogen). The normalized cDNA was ligated to λTriplEx2 vector (Takara), and packaged into recombinant lambda phages with Gigapack III plus (Stratagene, La Jolla, CA). The resulting library was used to develop plaques in XL1-Blue cells (Takara). The 1,118 developed plaques were identified, amplified by PCR with the pair of λTriplEx 5′ sequencing primers and the λTriplEx 3′ sequencing primer (Takara) and sequenced. A search for homologues of the obtained ESTs was performed by assembling sequences using SeqMan II (Dnastar, Madison, WI) and by BLAST searches (Altshul et al. 1997) against nonredundant database (blastn and blastx).

CAPS analyses

Based on the sequences of frequently occurring ESTs, or those with high E-values in BLAST searches, primers were designed using PRIMER 3 (http://www-genome.wimit.edu/cgi-bin/primer/primer3.cgi/). After screening for clean amplification from genomic DNA, the amplified fragments were sequenced and screened for polymorphism at restriction enzyme sites.

For CAPS analyses of the collected populations, the amplified fragments were digested with appropriate restriction enzymes and resolved by electrophoresis on 4% Nusieve 3:1 Agarose (Lonza, Rockland, ME).

SSR analyses

SSRs were identified using the Sputnik program (http://espressosoftware.com/pages/sputnik.jsp). Primers were designed to amplify fragments containing SSRs of longer than 16 bp using PRIMER 3. Seventeen primer sets designed within 16 ESTs were screened for clean amplification from genomic DNA and polymorphism.

Phylogenetic analyses based on sequences

Phylogenetic analyses were performed separately for chloroplast DNA sequences and nuclear GapCp sequences using the maximum parsimony (MP) method with PAUP* 4.0b10 (Swofford 2002). All characters were equally weighted and heuristic searches were conducted with 1,000 random addition replicates involving TBR branch swapping. Branch support was estimated by bootstrap analyses (Felsenstein 1985) with full heuristic searches, 1,000 bootstrap replicates, 10 random-addition-sequence replicates per bootstrap replicate, TBR branch swapping and MULTrees option on, and saving all trees.

Phylogenetic analyses based on genotypes

Genetic distances (Dps: the log-transformed proportion-of-shared-alleles distance) between all pairs of the examined individuals’ multi-locus genotypes of GapCp, six CAPS and three SSRs were calculated according to Bowcock et al. (1994). Based on the matrix of genetic distances, a neighbour-joining (NJ) tree was constructed using PHYLIP3.67 (Felsenstein 1981).

Population genetic analyses

Bayesian clustering was used to assess the relatedness among the examined individuals using the program STRUCTURE 2.2 (Pritchard et al. 2000). The number of populations (K) was estimated using the admixture ancestral model with correlated alleles, with K ranging from 2 to 7. Three independent runs of 50,000 MCMC (Markow Chain Monte Carlo) generations and 20,000 generations of “burn-in” were used for each value of K. The true number of populations is expected to be the value of K that maximizes the estimated model log-likelihood, log [P(X|K); Falush et al. 2003].

The posterior probabilities that the examined individuals fall into each of a set of hybrid classes (Parent 1, Parent 2, F1, F2, Backcross to Parent 1, Backcross to Parent 2, thus covering parents and two generations of offsprings) were also estimated by a Bayesian statistics-based method implemented in NewHybrids 1.0 (Anderson and Thompson 2002). The analysis was run for 20,000 MCMC repeats after a burn-in of 20,000 iterations.

Results

Morphology

Averages of the veinlet numbers of the basiscopic basal vein and the pinnule-base angles were plotted for each individual (Fig. 3). The two measured characters separated Osmunda lancea, O. × intermedia and O. japonica in Population H, their variation ranges did not overlap, and the variation range of O. lancea was narrower than that of O. japonica and O. × intermedia.

Fig. 3
figure 3

Morphological characters of pinnules in the individuals of Osmunda japonica (open circles), O. × intermedia (crosses) and O. lancea (solid diamonds). The individuals highlighted by squares indicate possible later-generation hybrids

MP trees based on chloroplast DNA and GapCp sequences

Nucleotide sequences of 3,661 bp were determined in chloroplast genome, and six haplotypes were identified. The single most parsimonious tree is shown in Fig. 4 with bootstrap percentages. Haplotypes a and d were found only in Osmunda japonica, and the haplotype of O. lancea of Kochi was not found anywhere else (Fig. 4). Haplotypes b, c and e were shared by O. lancea and O. japonica (Fig. 4). Thus, O. japonica and O. lancea were not distinct in the chloroplast DNA sequences.

Fig. 4
figure 4

Single most parsimonious tree based on chloroplast DNA sequences. Numbers in parentheses of terminal nodes are the numbers of individuals of each haplotype (J Osmunda japonica; L O. lancea). H, S and K indicate Hanno, Shojin and Kyushu populations, respectively. Asterisks correspond to individuals denoted by footnote a in Table 1. Numbers near branches and a number in parenthesis indicate nucleotide substitutions and bootstrap percentage, respectively

In contrast, O. japonica and O. lancea were distinct in the GapCp sequence. Nucleotide sequences of 300 bp were determined. For 55 out of 99 individuals, no heterozygous peak was observed in their electropherograms. Nucleotide substitutions were found at six sites in their sequences, and five alleles, La, Lb, Ja, Jb and Jc, were found. All the examined individuals of O. lancea were homozygotes of alleles La or Lb. Out of 30 individuals of O. japonica examined, 22 were homozygotes of alleles Ja, Jb or Jc. Out of 40 individuals of O. × intermedia examined, four individuals were homozygotes of allele La.

Heterozygous peaks were found in the electropherograms of 44 individuals. Those heterozygous peaks, however, can be explained as the results of combinations between two of the five already identified alleles, and their genotypes were inferable. Eight heterozygotes between alleles Ja and Jc were found in O. japonica; 36 individuals of O. × intermedia were heterozygotes between alleles Ja and La, Jb and La, or Jc and La.

The single most parsimonious tree based on the sequences of five alleles is shown in Fig. 5 with bootstrap percentages. The alleles of the same species were more closely related to each other than to those of the other species (Fig. 5).

Fig. 5
figure 5

Single most parsimonious tree based on GapCp sequences. Numbers in parentheses of terminal nodes are the numbers of homozygotes of each allele (J: Osmunda japonica; L: O. lancea). H, S and K indicate Hanno, Shojin and Kyushu populations, respectively. Numbers near branches and numbers in parentheses indicate nucleotide substitutions and bootstrap percentages, respectively

CAPS and SSRs analyses

Substitutional or insertion/deletion polymorphisms were observed in 26 fragments amplified with the primers designed to develop CAPS markers. There were polymorphisms at restriction enzyme sites in 6 of the 26 fragments. These six markers were clearly resolved upon electrophoresis. The primer sets to amplify these six fragments are listed in Table 2.

In the three amplified fragments containing SSRs, variations in repeat numbers were found in Osmunda japonica and O. lancea. The primer sets to amplify these three fragments are listed in Table 2.

Multi-locus genotypes of individuals

Only one genotype was found in Osmunda lancea of Population H, while there were 12 genotypes in O. japonica of Population H (Fig. 6). In Population H, O. lancea and O. japonica were entirely differentiated in five of the ten examined loci.

Fig. 6
figure 6

Neighbour-joining (NJ) tree based on genotypes in ten loci of nuclear genes. Asterisk, S and K indicate Hanno, Shojin and Kyushu populations, respectively. Genotypes highlighted by squares indicate possible later-generation hybrids

Taking all the collected individuals into consideration, O. lancea and O. japonica were entirely differentiated in three loci and have their own species-specific alleles (Table 2); the frequencies of alleles were considerably different in the other loci. Conspecific plants of O. lancea and O. japonica were clustered together in the NJ tree regardless of the collected localities (Fig. 6). In the plants of O. lancea, alleles were fixed in eight loci, and only three genotypes were found (Fig. 6). The three genotypes differ only in one or two loci of the ten examined loci. The observed heterozygosities of O. lancea were zero in all individuals, while those of O. japonica ranged from 0 to 0.4 (Fig. 7).

Fig. 7
figure 7

Frequency of heterozygosity. Open bars Osmunda japonica, grey bars O. × intermedia, solid bars O. lancea . Bars with numbers include possible later-generation hybrids. Line Expected frequency of heterozygosity assuming that all O. × intermedia plants are primary F1 hybrids between individuals of O. japonica and O. lancea

Osmunda × intermedia plants were positioned between the clusters of O. japonica and O. lancea in the NJ tree (Fig. 6). The O. × intermedia plants had the same alleles as those of O. japonica or O. lancea, and their genotypes can be explained as a combination of alleles of O. japonica and O. lancea. Comparison in multilocus genotypes of nuclear genes and chloroplast DNA haplotypes implies that 10 of 26 individuals of O. × intermedia in Population H are F1 hybrids between the pairs of O. japonica and O. lancea individuals sampled in this locality (Fig. 1).

The expected heterozygosities of F1 hybrids in Population H, based on the observed heterozygosities of the parental species, are shown in Fig. 7. The observed heterozygosities of O. × intermedia ranged from 0.4 to 1. Comparison of the observed heterozygosities of O. × intermedia with the expected heterozygosities of F1 hybrids shows that the number of individuals with the lower observed heterozygosities is larger than that with the expected values (Fig. 7).

Population structure

In all three independent STRUCTURE analyses, the highest likelihood was obtained for K = 3 clusters of plants. All of the 55 individuals from Population H were assigned to Osmunda lancea, O. japonica or O. × intermedia as identified by morphology (Fig. 8). At K = 2, O. × intermedia plants were shown to be admixtures of O. lancea and O. japonica (Fig. 8).

Fig. 8
figure 8

Population structure estimated by STRUCTURE analyses on the basis of data for ten nuclear gene loci obtained for 55 individuals. Each individual is represented by a single vertical bar divided into K different shades, where K is the number of populations assumed. Each shade represents one population, and the length of the shaded segment shows the individual’s estimated proportion of membership in that population

In a NewHybrids analysis, all of the 55 individuals from Population H were assigned to each species as identified by morphology (Fig. 9). Of the 26 with O. × intermedia morphology, one was identified to be an F2 hybrid with high certainty (>99.9%), and three were suggested to be F2 hybrids or backcross individuals with moderate probabilities (9–37%; Fig. 9). The other 22 plants of O. × intermedia from Population H and those from Population S and K were assigned to F1 hybrids with high certainty (>95%).

Fig. 9
figure 9

Probabilities assigned by NewHybrids analysis to each hybrid class on the basis of data for ten nuclear gene loci obtained for 55 individuals. Each individual is represented by a single horizontal bar divided into different hybrid classes

Discussion

In the nuclear genes examined, there was a remarkably large differentiation between Osmunda lancea and O. japonica (Fig. 6), although plants collected from geographically distant localities having similar genotypes (Fig. 6). In O. lancea, only three genotypes were found (Fig. 6), and its genetic diversity was very low (Fig. 7). Osmunda lancea is confined to streambeds in Japan, geographically isolated from those of different river systems, and their habitats are distributed patchily across their geographical distribution range. For that reason, it was unclear whether O. lancea originated monotopically or polytopically. The finding here that plants collected from geographically distant localities were genetically very similar does not suggest polytopic origins.

A noteworthy discrepancy between chloroplast and nuclear DNA phylogenies was revealed by the results of phylogenetic analyses. Osmunda lancea is not genetically distinct from O. japonica in the chloroplast DNA sequences determined (Fig. 4). This result is concordant with previous studies (Yatabe et al. 1999; Metzgar et al. 2008). The NJ tree based on the genotypes of nuclear genes marked by chloroplast haplotypes is shown in Fig. 10. Osmunda japonica and O. lancea share chloroplast haplotypes b and e, in spite of large genetic differentiation in their nuclear genes. The shared polymorphisms between species can be interpreted as resulting from introgression, and/or from the joint retention of ancestral polymorphisms (Gottlieb 1972; Heiser 1973; Rieseberg and Wendel 1993; Hey et al. 2004; Muir and Schlotterer 2005; Lexer et al. 2006; Patterson et al. 2006). In diverse plant taxa, apparent chloroplast DNA introgression, often without, or more extensive than, nuclear DNA introgression, have been reported (Rieseberg et al. 1996). A general explanation for the more rapid introgression of cytoplasmic markers compared to nuclear markers is the selection-linkage hypothesis (Martinsen et al. 2001; Funk and Omland 2003). In contrast to nuclear genes, cytoplasmic markers are usually unlinked to selectively disadvantageous alleles. Furthermore, the nuclear markers used in this study are all located in expressed genes, and possibly are under natural selection, which can maintain genetic differentiation irrespective of reproductive isolation. Although cytoplasmic introgression has not been reported in pteridophytes, low nuclear diversity and relatively high-chloroplast diversity in O. lancea can be explained by cytoplasmic introgression from O. japonica to O. lancea. Retention of ancestral polymorphisms in the chloroplast DNA sequences cannot be excluded, and further investigation is necessary in order to verify ongoing introgression between O. japonica and O. lancea.

Fig. 10
figure 10

Neighbour-joining tree based on genotypes in ten nuclear gene loci with chloroplast haplotypes. Numbers in parentheses indicate the number of individuals with the same genotypes

In Population H, less than half of the Osmunda × intermedia individuals were derived by hybridization between the individuals of O. japonica and O. lancea sampled in this locality (Fig. 1). It is possible that the immigration rate may be high, or that some of the parental individuals may have disappeared, because generation times are generally long in the genus Osmunda. All alleles found in O. × intermidia are, however, found either in O. japonica or O. lancea, and the genetic structure of O. × intermedia is an admixture of O. lancea and O.japonica (Fig. 8). This strongly supports the hypothesis that O. × intermedia is a hybrid swarm.

A few individuals of O. × intermedia had low heterozygosity. The heterozygosity of one individual was 0.4 and that of another was 0.6 (Fig. 7, individuals 1 and 3). Assuming that O. × intermedia consists only of primary F1 hybrids in Population H, the probability of the occurrence of individuals with heterozygosity of 0.4 and 0.6 is very low (Fig. 7). Those two individuals are suggested to be putative later-generation hybrids in the NewHybrids analysis, although the probability for individual 3 is not high (Fig. 9). Those individuals are considered to be F2 hybrids rather than backcross individuals as assigned in the NewHybrids analysis (Fig. 9), because they both have homozygotic loci with the alleles found only in O. japonica and those with the alleles fixed in O. lancea. The two individuals are morphologically different from the other O. × intermedia individuals (Fig. 3). One of them has a very broad pinnule-base, which is similar to O. japonica (Fig. 3, individual 1). The other has the bilateral symmetric pinnule base, similar to O. lancea (Fig. 3, individual 3). In O. japonica and O. × intermedia, the pinnule-base is unequal, and the basiscopic base generally truncate or rotund. The other two individuals assigned to F2 hybrids or backcross individuals in the NewHybrids analysis (Fig. 9, individuals 2 and 4) also have extreme morphology in O. × intermedia (Fig. 3). The morphological features of the four possible later-generation hybrids are considered to result from the segregation of genes for morphological characters. This may be the major reason why O. × intermedia in the H population has a large morphological variation. To conclude, O. × intermedia is a semifertile hybrid, as also suggested by previous morphological and cytotaxonomical studies (Shimura 1972; Shimura and Matsumoto 1977). The upper rheophytic zone, with less frequent flooding than the lower zone but more than the adjacent dryland, may be a hybrid zone across Japan, where O. × intermedia can survive.