Comparative plastome analysis of the Vandenboschia radicans complex in Korea: insights into its evolutionary history

Abstract

Background

Vandenboschia radicans and its close relatives are difficult to distinguish based on morphological characteristics because of reticulate evolution within the complex. Therefore, cytological and molecular approaches have been proposed for accurate identification of taxa within this complex. However, conventional plastid markers lack sufficient discriminatory power at the species level. In Korea, the V. radicans complex comprises nine genotypes associated with varying ploidy levels and hybrid origins. The maternal genotypes are classified into three groups: α, β, and γ. In this study, nine complete plastomes from the V. radicans complex were sequenced to investigate the evolution of different genotypes and find candidate plastome marker regions to facilitate the accurate identification of maternal lineage.

Results

The nine plastome sequences ranged from 146,846 to 147,247 bp, with identical gene content. The junctions between single copy and inverted repeat regions of Vandenboschia plastomes were categorized into three types. The phylogenetic tree inferred from 84 concatenated genes robustly supported the monophyly of each genotype within the V. radicans complex with high bootstrap values. However, the α genotype clade and its internal nodes showed only moderate gene concordance factors calculated from 84 individual gene trees. To identify candidate plastid markers for distinguishing maternal lineages, 1,719 phylogenetic trees were generated using sliding window analysis. Of these, 1,363 trees supported the monophyly of all genera within Hymenophyllaceae and all maternal genotypes within the V. radicans complex. Across the entire plastome, phylogenetic trees inferred from ycf2, rpoC2, ycf1, and rpoB showed almost identical topologies to the concatenated plastome tree based on the 84 concatenated genes. Among the 420 branch-model analyses (84 genes × 5 foreground branches), 20 yielded p-values below the significance threshold, and the estimated ω values for the foreground branches were less than 1, indicating purifying selection, rather than positive selection.

Conclusions

Short branch lengths in the phylogeny, purifying selection of plastome genes, minimal sequence length variations, and negligible genome rearrangements among plastomes of the V. radicans complex collectively indicate low genetic variation within the complex. Independent gametophyte stage, which allows populations to persist through clonal propagation, together with recent speciation was likely a major factor contributing to the low level of genetic variation within V. radicans complex.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12870-026-09583-7.

Background

Hymenophyllaceae Mart. represents an early diverging group of leptosporangiate ferns, comprising 434 species [1]. A distinctive feature of this family is its single-cell-layered leaves, which strongly support its monophyly. Despite this clear familial distinction, intrafamilial classification has been debated for many years [2–4]. Ebihara et al. [5] conducted a molecular revision of the Hymenophyllaceae, identifying nine genera and proposing morphological characteristics for the classification of infrageneric taxa. In the present study, we adhered to the generic circumscriptions defined by Ebihara et al. [5] for Hymenophyllaceae.

Vandenboschia Copel., as newly circumscribed by Ebihara et al. [5], comprises V. radicans (Sw.) Copel. and its relatives and is distributed across tropical and northern temperate regions [5]. Because of difficulties in distinguishing V. radicans from its close relatives, numerous cryptic taxa were historically included within V. radicans. For example, Ito [6] recognized three varieties of V. radicans in Japan based on frond size: V. radicans (Sw.) Copel. var. naseana (H. Christ) H. Ito, V. radicans var. orientalis (C.Chr.) H. Ito, and V. radicans var. nipponica (Nakai) H. Ito. These taxa were subsequently transferred to the genus Crepidomanes as C. radicans (Sw.) K. Iwats. var. naseanum (H. Christ) K. Iwats., C. birmanicum (Bedd.) K. Iwats., and C. amabile (Nakai) K. Iwats., respectively [7]. To ascertain species identities and elucidate reticulate evolution within the V. radicans complex in Japan and adjacent areas, Ebihara et al. [8] employed cytological and molecular approaches by chromosome counting, flow cytometric estimation of genome size, and analyses of the maternally inherited rbcL and biparentally inherited gapCp genes. Based on these data, the three varieties recognized by Ito [6] were reclassified as V. striata (D.Don) Ebihara, V. kalamocarpa (Hayata) Ebihara, and V. nipponica (Nakai) Ebihara, respectively. Interestingly, it was also revealed that hybrids and hybrid species of this complex, including amphidiploids, have wider distributions than their diploid progenitors. Ebihara et al. [8] did not found the evidence of apogamous reproduction within the complex and suggested the recent origin of many hybrids based on field observations. They further proposed that stress-tolerant independent gametophytes may serve as a persistent source of gametes, thereby facilitating recurrent hybridization.

In Korea, the V. radicans complex has been known to comprises two species (V. nipponica and V. kalamocarpa) and two hybrids (V. x quelpaertensis (Nakai) Ebihara and V. x stenosiphon (Christ) Copel.) based on their morphological characteristics [9]. In contrast to the distribution patterns observed in Japan, where triploid hybrids are widespread whereas non-hybrid diploids are restricted to limited habitats [8], both non-hybrid species and hybrids were sampled in similar proportions in Korea, based on the records of herbarium specimens. To achieve precise species identification within the complex, Park et al. [10] genotyped Vandenboschia individuals from Jeju Island and Ulleung Island, Korea, using rbcL and gapCp, as recommended by Ebihara et al. [8]. They also verified the ploidy levels of these individuals by chromosome counting and flow cytometry. The results revealed nine genotypes associated with varying ploidy levels and hybrid origins. Notably, they identified an unrecorded hybrid in Korea: V. kalamocarpa × V. nipponica × V. striata [10]. Although their findings largely aligned with those of Ebihara et al. [8], the reproductive capabilities of the hybrids remain inconclusive. Ebihara et al. [8] observed sterile diploid and triploid hybrids with irregular meiosis and spore formation. In contrast, Park et al. [10] successfully grew a small number of gametophytes from triploid hybrids over 2 years, suggesting that dominant hybrids may provide a limited supply of viable gametophytes.

Plastomes in vascular plants range from 120 to 170 kb in length and contain approximately 110–130 genes [11–13]. Their highly conserved nature makes them valuable molecular markers for plant phylogenetic studies [14–16]. The rbcL gene, in particular, has been instrumental in resolving phylogenetic relationships among ferns [17–19]. An ENTREZ search based on plastome gene names [20] with sequence lengths ranging from 500 to 5,000 bp in Polypodiopsida revealed that only a few molecular markers have been used disproportionately to date (Fig. 1). The rbcL gene accounted for approximately 38% of the results, and the top five genes (rbcL, rps4, trnL, atpB, matK) collectively accounted for 78%. Although this ENTREZ search did not distinguish partial sequences, such as the 5′ rbcL region, 3′ rbcL region, or the rbcL-atpB intergenic spacers (IGSs), all were retrieved as rbcL. Nevertheless, the strong dependence of rbcL in Polypodiopsida is clear. However, the rbcL gene alone often lacks sufficient discriminatory power at the species level [21]. This limitation is evident in the V. radicans complex, where V. kalamocarpa and V. subclatrata K. Iwats share the same maternal rbcL genotype, despite exhibiting distinct morphology and distribution [8].

Fig. 1.

Nucleotide hits retrieved from GenBank for Polypodiopsida using plastome gene names and sequence lengths via an ENTREZ search.

In the present study, we sequenced the complete plastomes of nine genotypes of the V. radicans complex collected from Korea to 1) thoroughly investigate the evolution of different genotype plastomes and 2) identify candidate plastome marker regions to facilitate the accurate identification of maternal lineages.

These objectives aimed to enhance our understanding of genetic diversity and evolutionary relationships within the V. radicans complex, thereby providing a foundation for more precise species identification and robust phylogenetic studies.

Materials and methods

Plant materials and DNA extraction

To analyze the complete plastome of the V. radicans complex distributed in Korea, we used plant samples from a previous study [10] which were transplanted into a greenhouse at Chungbuk National University after collection from the natural habitats of Jeju and Ulleung Islands. The genotypes of the plants were determined by Park et al. [10] and refined them referring the terms described by Ebihara et al. [8], using the symbols α, β, γ, and * (to indicate unknown). The first characteristic of the genotype is the maternal genome type.

Fresh leaves from the nine different genotypes were used for DNA extraction. Genomic DNA was extracted using the DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s protocol.

Plastome sequencing, assembling, and annotation

High-quality DNA samples from the nine genotypes were used to prepare next-generation sequencing libraries, which were sequenced by Macrogen (Seoul, Korea) using a HiSeq X Ten platform (Illumina, San Diego, CA, USA). Raw reads were trimmed using Trimmomatic 0.36 [22] with the following options: leading, 10; trailing, 10; sliding window, 4:20; and minlen, 50. The assembly followed the methodology described by Kim et al. [23]. Trimmed reads were aligned to previously sequenced V. speciosa (NC_041000) using the Geneious assembler with medium–low sensitivity in Geneious prime [24]. Assembled reads were then de novo assembled using the Geneious assembler, with zero mismatches and gaps allowed among reads. The reads were then re-aligned to the resulting de novo contigs with zero mismatches and gaps, with 100 iterations. Finally, the resulting contigs were then de novo assembled, using the circularize contigs and matching ends options. After trimming, the numbers of remaining reads, plastome-mapped reads, and the average coverage depth for each sample were summarized in Supplementary Table 1.

Gene annotation was performed by comparing the assembled sequences with V. speciosa genes using Geneious Prime [24]. The annotation was further validated using a BLAST search [25] and tRNAscan-SE [26]. The ReFernment R package was used to annotate the putative RNA editing sites in genes expected to be prevalent in leptosporangiate fern plastomes [27]. To assess inverted repeat (IR) contraction and expansion in Vandenboschia plastomes, we analyzed the junctions between single-copy (SC) and inverted repeat regions (IRs) using CPJSdraw [28].

Phylogenetic analyses

To investigate phylogenetic relationships within the V. radicans complex, 12 related plastome sequences were downloaded from GenBank. These included three sequences from Vandenboschia as ingroups and one from Callistopteris Copel. Two from Crepidomanes C. Presl, five from Hymenophyllum Sm., and one from Trichomanes L. were used as outgroups (Supplementary Table 2). The maternal genotypes of V. speciosa (Willd.) G.Kunkel were designated as the “Spe genotype” to distinguish it from the other plastome genotypes found in Korea.

Eighty-four coding genes were extracted from 21 plastome sequences within Hymenophyllaceae and aligned using MUSCLE [29]. Partition models were selected using ModelFinder with edge-proportional partitions and merge options [30]. A phylogenetic tree was constructed using the maximum likelihood (ML) method with IQ-TREE3 [31], incorporating 10,000 ultrafast bootstrap replicates [32, 33]. The resulting ML tree was treated as the concatenated plastome tree for subsequent analyses. The gene concordance factor (gCF) for the 84 coding genes was calculated to determine the gene support frequency for each branch of the concatenated plastome tree using IQ-TREE3 [31, 34].

To investigate the candidate markers that were topologically similar to the concatenated plastome tree, a sliding window analysis was performed. Twenty-one plastome sequences from Hymenophyllaceae were aligned using MAFFT [35] and subsequences were extracted with a window size of 1,000 bp and a step size of 100 bp using Biostrings package [36]. If any taxon contained gaps only within a given region, the window was skipped. Tree distances from concatenated plastome tree to these gene trees were calculated using the Kendall-Colijn metric [37] within the treespace package [38].

Pairwise genetic distance among the plastomes within the V. radicans complex

Uncorrected p-distances among plastomes within the V. radicans complex were calculated using the DECIPHER package [39]. Self-comparisons and duplicate pairwise comparisons were removed. Genotype information was then merged with the pairwise distance data to classify comparisons as either within-genotype or between-genotype. Minimum, maximum and mean p-distances were calculated for each category.

Distribution of simple sequence repeats and insertions/deletions

To investigate the distribution of simple sequence repeats (SSRs) in the plastome across the genera of Hymenophyllaceae and the genotypes of the V. radicans complex, 21 plastome sequences (including 12 downloaded sequences) were analyzed using MISA [40] with default settings.

To investigate insertions and deletions (indels) within the genus Vandenboschia, 12 plastome sequences (Supplementary Table 2), including three downloaded Vandenboschia sequences (V. auriculata, V. speciosa, and V. striata), were aligned using MAFFT [35]. However, the plastome of V. auriculata caused alignment ambiguities because of its high genetic distance from the V. radicans complex, leading to irregular indels within the genus Vandenboschia. Manual sequence alignment adjustments were required to improve the accuracy of the indels. Therefore, indels were analyzed only within the V. radicans complex, excluding V. auriculata. Eleven plastome sequences of V. radicans complex were aligned using MAFFT [35]. The resulting multiple sequence alignment was further refined via sliding window analysis with a 500 bp window size and 200 bp step size using MUSCLE [29] to improve the accuracy of indel detection within intergenic spacers. All gaps, represented by ‘-’ in the multiple sequence alignment, were detected and grouped by taxon. An indel was classified as a genotype-specific indel when it was detected in all plastomes of one genotype but was absent from all plastomes of the other genotypes. Some genotype-specific indels overlapped when large indels encompassed the small indels.

Small inversions within the V. radicans complex

Palindromic sequences were identified from multiple sequence alignments of 11 plastome sequences of the V. radicans complex. The detection criteria included a minimum stem length of 3 bp, a minimum width of 8 bp, and loop lengths ranging from 2 to 50 bp. Of these palindromic sequences, small inversions were detected when the stem positions were identical; however, at least two sequences at a given stem position were reverse complements of each other.

Selective pressure analyses

Eighty-four protein-coding genes were aligned at the codon level using MACSE v2 [41]. The resulting alignments and concatenated plastome tree described in the “Phylogenetic analyses” section were used to investigate selective pressure across Vandenboschia. Based on maternal genotypes within the V. radicans complex, we defined five foreground branches, namely, α genotype, β genotype, γ genotype, V. speciosa, and Vandenboschia clades. Codeml in PAML [42] was used to compare the one-ratio model (M0), which assumes a single ω (dN/dS) value across the entire tree, against a two-ratio branch model that allows a positive ω value to differ between foreground and background branches. The significance level was set to an alpha value of 0.05, and p-values were adjusted using the Bonferroni correction [43] and the Benjamini–Hochberg procedure [44] to control for multiple testing of a gene.

Data analyses and visualization

All data were analyzed using R 4.4.1 [45] with the following packages: DECIPHER [39], dplyr [46], Biostrings [36], and Stringr [47]. The packages treeio [48], ggtree [49], and ggplot2 [50] were used to visualize the data and plot objects.

Results

The plastome sequences of the V. radicans complex distributed in Korea

The lengths of the nine plastome sequences range from 146,846 bp for the β genotype of V. x stenosiphon to 147,247 bp for the γ genotype of V. kalamocarpa × V. nipponica × V. striata (Table 1). Length variations among genotypes were more pronounced in large single-copy (LSC) regions, followed by IR and small single-copy (SSC) regions.

Table 1.

Information on plastome sequences of the Vandenboschia radicans complex in Korea

Species	Voucher specimena	Ploidy	Genotypeb	Total length	LSC	SSC	IR	Accession
V. kalamocarpa	CBNU2018-0156	2x	αα	147,122	89,821	20,903	18,199	PQ874236
V. kalamocarpa	CBNU2023-0007	3x	ααα	147,154	89,849	20,903	18,201	PQ874244
V. x stenosiphon	CBNU2018-0273	2x	αβ	147,123	89,824	20,903	18,198	PQ874238
V. x quelpaertensis	CBNU2020-0131	3x	ααγ	147,128	89,814	20,904	18,205	PQ874241
V. kalamocarpa × V. nipponica × V. striata	CBNU2018-0377	4x	ααβγ	147,139	89,838	20,903	18,199	PQ874239
V. x stenosiphon	CBNU2018-0160	3x	βα*	146,846	89,632	20,882	18,166	PQ874237
V. nipponica	CBNU2022-0064	2x	ββ	146,998	89,768	20,898	18,166	PQ874243
V. nipponica	CBNU2019-0032	3x	βββ	146,998	89,755	20,899	18,172	PQ874240
V. kalamocarpa × V. nipponica × V. striata	CBNU2022-0008	3x	γαβ	147,247	90,104	20,861	18,141	PQ874242

Voucher specimen ^a = herbarium voucher code generated by Park et al. [10]; Ploidy = chromosome ploidy level; Genotype ^b = combined nuclear and plastid genotype (α, β, γ represent specific haplotypes, * indicates unknown, underline denotes maternal genome type); Total length = total plastome size (bp); LSC = large single-copy region length (bp); SSC = small single-copy region length (bp); IR = inverted repeat region length (bp); Accession = GenBank accession number

The gene content among the nine plastomes was identical, with 129 genes identified, including 87 coding genes, 34 tRNAs, and eight rRNAs. Among these, three coding genes, five tRNAs, and four rRNAs were duplicated in the IR region. Thirteen genes had one intron (atpF, ndhA, ndhB, petB, petD, rpl2, rpoC1, rps16, trnA-TGC, trnG-TCC, trnI-GAT, trnL-CAA, trnV-TAC) and three genes had two introns (clpP, rps12, ycf3). The intron of rpl16 found in the V. auriculata plastome was absent from all Korean V. radicans complexes, as well as V. striata and V. speciosa.

The junctions between SC/IR of Vandenboschia plastomes were categorized into three groups: group I comprising the α, β, and Spe genotypes of the V. radicans complex; group II comprising the γ genotype of the V. radicans complex; and group III comprising V. auriculata (Supplementary Fig. 1). These structural mutations were caused by SC/IR junction shifts to SC regions. Compared to group I, the IRa/LSC junction in group II shifted by 9 bp into the LSC region, whereas in group III, it shifted by 20 bp. Similarly, the IRb/SSC junction in group II shifted 1 bp into the IRb region, whereas in group III, it shifted 23 bp compared with that in group I.

The pairwise genetic distances within and between the genotypes of the V. radicans complex are shown in Table 2. The average p-distance within the γ genotypes was approximately four times higher than that within the α and β genotypes. The maximum p-distances among the α, β, and Spe genotypes ranged from 0.0078 to 0.0110, whereas the minimum p-distances between the γ genotypes and the α, β, and Spe genotypes ranged from 0.0181 to 0.0196.

Table 2.

Pairwise genetic distances within and between genotypes of the V. radicans complex

Comparisona	Mean	Min	Max
Within the α genotype	0.0005	0.0004	0.0006
Within the β genotype	0.0005	0.0003	0.0007
Within the γ genotype	0.0020	0.0020	0.0020
Between α and β genotypes	0.0077	0.0076	0.0078
Between α and γ genotypes	0.0198	0.0196	0.0200
Between α and Spe genotypes	0.0109	0.0108	0.0110
Between β and γ genotypes	0.0186	0.0184	0.0188
Between β and Spe genotypes	0.0096	0.0096	0.0097
Between γ and Spe genotypes	0.0183	0.0181	0.0184

^a The Spe genotype had one plastome sequence; therefore, within-genotype comparisons are not available

Phylogenetic position of the V. radicans complex based on the plastome sequence

Based on the concatenated plastome tree, the genus Vandenboschia was found to be monophyletic and closely related to Crepidomanes. Within Vandenboschia, V. auriculata was identified as a sister to the V. radicans complex, with a strongly supported bootstrap value (Fig. 2). The monophyly of each genotype within the V. radicans complex, namely α, β, γ, and Spe, was also robustly supported by high bootstrap values, despite Spe being represented by only a single individual. Genotypes α and β form a clade and are closely related to the Spe genotype. The γ genotypes were identified as the sister group to the others of the V. radicans complex.

Fig. 2.

Maximum likelihood phylogenetic tree based on the 84 coding genes. Numbers on the nodes (upper-left box) indicate bootstrap support values. Branch colors represent different maternal genotypes in Vandenboschia

Using the concatenated plastome tree, concordance with 84 individual gene trees was analyzed (Supplementary Fig. 2). The clade of the α genotypes was supported by a gCF of 69% and the internal nodes within this clade also showed weak support. In contrast, the clade of β genotypes was strongly supported with a gCF of 90.5%. The gCFs among genotypes were moderate (75% and 89.3%) except for the crown group of the α and β genotypes, which showed weak support (58.3%).

Candidate plastid markers

A total of 1,719 phylogenetic trees were generated using sliding window analysis. The distance between each tree and concatenated plastome tree ranged from 0 (windows 66,601–67,600: ndhJ and ndhJ ~ ndhK) to 77 (windows 124,601–125,600: trnV-GAC ~ rrn16), with a mean distance of 12.5. Of the 1,719 trees, 1,363 supported the monophyly of all genera in Hymenophyllaceae and all genotypes within the V. radicans complex, regardless of the relationships among these genera or genotypes (Supplementary Fig. 3). To identify candidate molecular marker regions, we counted the numbers of window across genes and IGSs and then plotted genes and IGS regions longer than 500 bp (Fig. 3). A total of 37 genes and 11 IGSs contained more than one informative window; notably, ycf2, ycf1, rpoC2, and rpoB contained more than 20 windows (Fig. 3A). The number of informative windows per kb was highest for ycf2, followed by rpoC2, rpoB, and ycf1 (Fig. 3B).

Fig. 3.

Point plot showing the number of windows within each region. Red and blue points represent genes and intergenic spacers, respectively

Selective pressure on each plastome gene across Vandenboschia

Among the 420 branch-model analyses (84 genes × 5 foreground branches), 20 yielded p-values below the significance threshold, involving 19 genes (Supplementary Table 3). However, after adjustment using the Bonferroni correction and the Benjamini–Hochberg procedure, only three analyses (γ genotype clade on the psaA gene tree; Vandenboschia clade on the psbB gene tree; Vandenboschia clade on the rps3 gene tree) remained statistically significant. In addition, in all 20 models, the estimated ω values for the foreground branches were less than 1, indicating purifying selection rather than positive selection.

Distribution of SSRs and indels

Mono-, di-, tri-, and tetra-nucleotide repeats were distributed across the plastomes of Hymenophyllaceae, with mononucleotide repeats being the most abundant (Supplementary Fig. 4). In Vandenboschia, dinucleotide and trinucleotide repeats were found only in V. auriculata but not in the V. radicans complex.

In total, 1,530 indels were identified within V. radicans complex plastomes. Among them, 232 loci were genotype-specific indels (Fig. 4), with 31 unique to the α genotype, 23 unique to β, 38 unique to Spe, and 111 unique to γ. Although the α and β genotypes were closely related based on concatenated plastome tree, they shared only 18 indels. Indels were also found between paraphyletic genotypes pairs, with six shared by α and Spe or β and γ, and five by α and γ or β and Spe.

Fig. 4.

Genotype-specific indels in the plastomes of the Vandenboschia radicans complex. The numbers and + on the x axis indicate the length and a length greater than 10 bp, respectively

Palindromic sequences and small inversions within the V. radicans complex

In total, 1,172,618 palindromic sequences, including those with loops longer than 2 bp, were found across the 11 V. radicans complex plastome sequences, with an average of 106,602 per plastome. The frequency remained constant, regardless of the increase in loop length (Supplementary Fig. 5), whereas the frequency and stem length were inversely logarithmic (Supplementary Fig. 6).

Five small inversions were detected across plastomes of the V. radicans complex (Supplementary Fig. 7). The lengths of the stems and loops of the palindromic sequences exhibiting small inversions ranged from 3 to 21 bp and 2 to 6 bp, respectively. Four of these inversions consisted of two types of loops with reverse complementary relationships, whereas the other showed additional substitutions and sequential inversions.

Discussion

The phylogenetic relationships within V. radicans complex in Korea

The reticulate evolution of the V. radicans complex in East Asia has been well described by Ebihara et al. [8]. They counted chromosome numbers for representative taxa and measured genome sizes of the remaining taxa by flow cytometry to determine the ploidy levels of the V. radicans complex, and sequenced nuclear and plastid genes to confirm hybridization among non-hybrid species. In phylogenetic studies using the rbcL gene alone [8, 51, 52], including the rbcL tree reconstructed in this study (identical in topology to those of previous studies; figure not shown), α genotypes and V. speciosa formed a clade, with β genotypes as their closely related sister group. In contrast, the phylogenetic tree based on 84 concatenated plastid genes strongly supported a close relationship between the α and β genotypes, as well as the monophyly of each genotype (Fig. 2). To fully understand the discrepancy between the phylogenetic tree based on 84 concatenated plastid genes and the rbcL tree, the evolutionary history of V. kalamocarpa, V. nipponica, and V. speciosa should be considered.

Based on cytological and molecular data [8], Trichomanes kalamocarpum and T. nipponicum were transferred to the genus Vandenboschia, resulting in V. kalamocarpa and V. nipponica, respectively [53]. V. kalamocarpa is distributed across East Asia [54], whereas V. nipponica is just found in Korea and Japan [10]. Although these two species were previously thought to have broad distributions based on morphological identification, recent studies using molecular data have indicated that their distributions are more restricted than previously recognized [10, 53]. In contrast, V. speciosa is a Macaronesian-European endemic species [55] and is considered tetraploid [56, 57]. A recent study of V. speciosa populations revealed two gapCp variants. One variant formed a clade with V. kalamocarpa, V. nipponica, and one of the two gapCp sequence types found in V. boschiana J.W.Sturm ex Bosch, corresponding to the α and β genotypes, whereas the other formed a clade with V. striata, representing the γ genotype [52]. Based on maternally inherited molecular markers, it was revealed that V. speciosa was the closest relative to V. boschiana based on rbcL, whereas V. boschiana was nested within V. speciosa based on trnH-psbA. The gapCp sequences of V. boschiana also supported its hybrid origin [52], with V. kalamocarpa or the common ancestor of V. kalamocarpa and V. nipponica serving as the maternal progenitor. If the topology inferred from the previous rbcL tree is accepted, V. kalamocarpa and V. nipponica first diverged. Then, V. boschiana originated through hybridization between V. kalamocarpa and V. davallioides. Subsequently, V. boschiana may have served as the maternal parent of V. speciosa. However, the α-type gapCp sequences of V. speciosa were sister to the clade comprising V. kalamocarpa, V. nipponica, and V. boschiana. In addition, p-distance analyses showed that the α and β genotypes were more similar to each other than either was to the Spe genotype. Taken together, it is likely that V. speciosa originated from the hybridization between an ancestor of α and β genotype and a γ genotype, subsequently spreading to its current ranges. Consequently, phylogenetic inference based solely on the rbcL gene may be limited by its low phylogenetic resolution in Vandenboschia, particularly because of frequent hybridization.

Newly proposed plastid markers for Vandenboschia phylogeny

Among the plastid nucleotides ranging from 300 to 3,000 bp in GenBank, the rbcL sequence is the dominant marker used for three genera: Hymonophyllum, Trichomanes, Vandenboschia (some members of the V. radican complex are still listed under Trichomanes in GenBank). rps4-trnS, rbcL-accD, atpB, trnR-trnG, matK, and trnL-trnF are also used as molecular markers, followed by rbcL in Hymenophyllum. However, in Trichomanes and Vandenboschia, IGS regions, such as psbA-trnH, trnL-trnF, atpB, and rps4 have been limited and primarily applied in DNA barcoding [58, 59], haplotype analysis of single species [52], and studies of deep phylogeny [60, 61].

The discrepancy between the concatenated plastome tree and gene trees also indicates that these conventional maternal gene markers are no longer useful, at least in resolving Vandenboschia phylogeny. The nodes representing the α and β genotypes are strongly supported in the concatenated plastome tree with 100 bootstrap value (Fig. 2); however, only 69% and 70.2% of the gene trees support these nodes, respectively (Supplementary Fig. 2). Similarly, the clade comprising the diploid V. nipponica and triploid V. x stenosiphon received 100 bootstrap supports; however, only 19% of the gene trees supported this node. Low gCF values are generally interpreted as reflecting either true discordance among individual gene trees or weak phylogenetic signals at the corresponding loci [32].

Tree-to-tree distances between the gene trees and the concatenated plastome tree based on sliding window analysis (Fig. 3) indicated that ycf2 contains numerous windows that can be used as molecular marker regions. In particular, all 52 windows within ycf2 in the MSA of Hymenophyllaceae passed the monophyly test of groups that included more than two taxa (Supplementary Table 4). The high variability in ycf2 in ferns is not a novel finding. This gene has previously been shown to be highly diverse in plastome analyses of Cyatheaceae [62] and Equisetaceae [63], and has been used as a molecular marker to delimit Asplenium scolopendrium complex, where universal markers fail to distinguish them reliably [64].

The origin of the V. radicans complex in Korea

Specimens of the V. radicans complex collected from inland Korea have been deposited in the national herbaria [10]. However, most extant populations are found on Ulleung Island, an oceanic island noted for its high level of anagenetic speciation [65], and on Jeju Island, located on the East China–Japan land bridge during glacial periods [66].

In this study, two non-hybrid β genotype individuals from Ulleung Island and one hybrid β genotype individual from Jeju Island were analyzed. In the concatenated plastome tree, the diploid V. nipponica (Ulleung Island) is a sister to the triploid V. x stenosiphon (Jeju Island) and forms a clade with the triploid V. nipponica (Ulleung Island). The genetic distance between diploid V. nipponica and triploid V. x stenosiphon is approximately half of that between the two non-hybrid β genotype individuals from Ulleung Island. These relationships are strongly supported by bootstrap values, in contrast to those of the α genotypes (Fig. 2). Based on these results, we propose two hypotheses to explain the phylogenetic relationships of them.

The first is the self-supply hypothesis. The V. radicans complex occurs in highly humid, windless hollows on both Ulleung and Jeju Islands, conditions that are generally unfavorable for wind-mediated spore dispersal. The spores of Hymenophyllaceae are chlorophyllous [67] have a lack of dormancy and short viability [68]. Therefore, establishing populations via inter-island dispersal would require the continuous dispersal of large numbers of spores from large populations to serve as maternal parents in hybridization events. However, the only small populations containing di- and triploids of β genotypes are currently present on both Ulleung and Jeju Island [10]. This suggests that much larger populations existed on one or both islands, enabling continuous long-distance dispersal in the past, and that these populations have recently contracted to their present size.

The second is the external supply hypothesis. The gapCP composition of the β genotypes in Korea is nearly identical to that in Japan, except for one Korean-specific variant [10]. In Japan, β genotypes are primarily distributed along the eastern coast, and the number and size of populations is relatively larger than in Korea [54]. The larger populations in Japan may produce more spores, thereby increasing the likelihood of long-distance dispersal in Korea. As a result, annual supplementation in Korea with newly dispersed spores from Japan may be possible, although their viability cannot be guaranteed. These two hypotheses are important for developing a conservation strategy for Vandenboschia species in Korea. V. nipponica has been listed as endangered on the Korean Red List. According to Mr. S. H. Park’s personal comments, most Vandenboschia populations are hybrid V. x stenosiphon, although many specimens have been identified as non-hybrid Vandenboschia species. The first hypothesis suggests that currently existing populations have also experienced a decline in size because of environmental stress. Therefore, they should be protected by strong efforts, and ex situ conservation should be considered immediately. However, if there are external supplies, we need to focus on the types of spores in Vandenboschia species that are dispersed in Korea and where they originate. Further studies are needed to confirm these hypotheses with expanded sampling across their distributions, especially in Japan.

Length polymorphism of plastomes in the V. radicans complex

Microsatellites have been tested as molecular markers for plant genetics [69] and have been widely applied across plant science [70–72], including for Vandenboschia species [73]. Furthermore, the increasing availability of whole-genome sequences driven by the rapid development of high-throughput sequencing techniques has facilitated the identification of informative SSRs [74]. Similarly, polymorphic SSRs in plastomes have been shown to be highly informative in plant population studies [75] because of their relatively high levels of intraspecific variability [76]. Although there are some limitations associated with cpSSRs, such as size homoplasy, heteroplasmy, and cytoplasmic introgression [77], the conserved structure of plastomes in land plants facilitates primer design and broadens the applicability of cpSSRs [78–80].

The SSR distribution in the plastomes of Hymenophyllaceae showed that four genera, excluding Vandenboschia, contained more than three types of repeat units, with the exception of Callistopteris, for which only a single plastome has been sequenced (Supplementary Fig. 4). Within the genus Vandenboschia, the V. radicans complex, belonging to the subgenus Vandenboschia, contains only mononucleotide repeats, whereas V. auriculata, a member of the subgenus Lacosteopsis and considered part of the basal lineage in Vandenboschia [81], possesses mono-, di-, and trinucleotide repeat units. In addition, all branch lengths among genotypes within the V. radicans complex were shorter than those observed at the species level in other genera (Fig. 2). These observations suggest that SSR abundance and diversity may have decreased in the ancestor of the V. radicans complex following its divergence from the subgenus Lacosteopsis. Consequently, the relatively short duration of speciation within the complex may not yet allowed the accumulation of additional SSR diversity.

Of the indels within the V. radicans complex, 84.8% were not associated with any specific lineage. Most genotype-specific indels occur in a single genotype, and genotype-specific indels longer than 10 bp are primarily located in the plastomes of the γ genotypes (Fig. 4). In particular, large indel-related enlarged non-coding regions of fern organelles (ENRFOs), which can be used as molecular markers for the identification of closely related taxa in ferns [13, 82, 83] were not found in the plastomes of V. radicans complex. These results suggest that indel mutations occur following speciation within the complex, and that they may be subject to selective pressure.

It was unable to test the availability of these SSRs and indels at the population level of V. radicans complex in Korea because of the limited number and small size of available populations at the time [10]. However, hybridization within V. radicans complex appears to occur frequently without fixation of the maternal parent. These length polymorphisms, SSRs, and indels may provide valuable insights into the evolution of this morphologically indistinguishable complex, particularly regarding maternal lineage.

Structural evolution of plastomes among different genotypes within the V. radicans complex

The structural features of the plastome of each genotype were distinguished. The SC/IR junctions of Vandenboschia plastomes were categorized into three groups, consistent with their phylogenetic relationships (Supplementary Fig. 1). However, in this genus, only a few nucleotides are transferred across the SC/IR junctions, whereas in the genus Hymenophyllum, all genes are relocated in or out of the SC/IR junctions [82]. ENRFOs found in the genera Hymenophyllum [82] and Crepidomanes [84] were not detected. ENRFOs were first identified in the plastome of Mankyua chejuense, an endemic species on Jeju Island, as foreign DNA from cyanobacteria or basal early diverging photosynthetic organisms [13]. These peculiar DNA sequences have been found in at least one species per genus in ferns for which several plastome sequences have been reported [82] but not in the genus Vandenboschia.

Consequently, the plastomes of the V. radicans complex were highly conserved in structure; however, it was possible to distinguish the genotype within the complex based on differences in the SC/IR junctions. The absence of ENRFOs, large indels, and gene rearrangements also suggests that they are under selective pressure in the structure.

Low genetic variations affected by independent gametophytic life in the complex

Hybrids and hybrid polyploids in the complex distribute more widely than diploids in both Korea and Japan [8, 10]. Ebihara et al. [85] found independent gametophytes that seem to be maternal parents for hybrids where diploid sporophytes are absent or removed distantly, and they suggested the hypothesis of “hybridization involving independent gametophytes.” Although certain hybrid spores of the V. radicans complex seem viable [10], we agree with this hypothesis based on our observations. Independent gametophytes grow slower than annual cordiforms [86] and have the capacity to spread by air and form a new colony with asexual reproduction [87]. Therefore, the generation length of independent gametophytes is meaningless and is correlated with the rates of molecular evolution [88]. If these independent gametophytes are the major suppliers of generative cells (one or both) for diploids and their hybrids, the rate of evolution may be very slow.

Conclusions

Nine complete plastomes representing the nine genotypes of the Korean V. radicans complex, which differ in ploidy levels and hybrid origins, were sequenced to investigate the evolution of these genotypes and identify candidate plastome marker regions for accurate identification of maternal lineages. Based on the phylogenetic tree constructed using 84 genes, the monophyly of each genotype within the V. radicans complex was robustly supported by high bootstrap values. The topologies of ycf2, rpoC2, ycf1, and rpoB were almost identical to concatenated plastome tree, and these genes were proposed to be candidate plastome marker regions to trace maternal lineages of this complex. Short branch lengths in the phylogeny, purifying selection of plastome genes, minimal sequence length variations, and negligible genome rearrangements among plastomes of the V. radicans complex collectively indicate low genetic variation within the complex. Independent gametophyte stage, which allows populations to persist through clonal propagation, together with recent speciation was likely a major factor contributing to the low level of genetic variation observed in V. radicans complex.

Authors’ contributions

Conceptualization, DTV, HTK, and JSK; methodology, HTK; data curation, HTK; resources, HTK and JSK; writing-original draft preparation, HTK; writing-review and editing, DTV, HTK, and JSK; visualization, HTK; project administration, JSK; and funding acquisition, HTK and JSK. All the authors have read and approved the final version of this manuscript.

Funding

This research was supported by the Gyeongbuk ANCHOR system through the Gyeongbuk ANCHOR CENTER, funded by the Ministry of Education(MOE) and the Gyeongsangbuk-do, Republic of Korea. (B0080319004414).

Data availability

The plastome data generated and analyzed in the current study are available in the NCBI database under the accession numbers listed in Table 1. For example, the plastome data of diploid V. kalamocarpa with αα genotype is accessible via the link https://www.ncbi.nlm.nih.gov/nuccore/PQ874236.1/.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.