We compared the morphological characters of G. crassifolia, G. schlechtendaliana, G. × tamnaensis, and G. similis from herbarium specimens deposited in AICH, HIBG, HYO, KYO, MAK, SCM, TI, and TNS and from living plants collected throughout Japan during fieldwork between 2011 and 2021. Morphological variations among G. schlechtendaliana, G. × tamnaensis, and G. similis were further investigated by reviewing the literature. Morphological characters were visually observed under a Leica M165C stereomicroscope and measured using a digital caliper. The dissected floral parts were photographed using an Olympus OM-D E-M1 Mark II digital camera equipped with an Olympus 30 mm macro lens or a Leica MC170 HD digital camera attached to a Leica M165C stereo microscope. Since we revealed that G. crassifolia is distributed widely throughout Japan, we also provided a revised description of G. crassifolia based on the newly discovered specimens from our field surveys and herbarium investigations. At least one voucher specimen from each new population discovered during our field survey was deposited in KYO and TNS (Suppl. material 1). The herbarium acronyms follow Index Herbariorum (Thiers 2022).

Root tips were collected from five individuals of G. crassifolia (representing five populations) and four individuals of G. schlechtendaliana (including a G. schlechtendaliana var. yakushimensis Suetsugu & H.Hayak. individual; representing three populations). They were used for mitotic chromosome counts, as described in Suetsugu et al. (2019). Root tips were pretreated with 2 mM 8-hydroxyquinoline solution for 4–5 h, fixed in Carnoy’s solution for 1–24 h, macerated in 1 N HCl at 60 °C for 1 min, and then squashed in aceto-orcein. The samples were then observed and photographed under a light microscope.

The breeding systems of G. schlechtendaliana and G. crassifolia were investigated during early-to-late September 2016 in a sympatric population in Kami-shi, Kochi Pref., Japan. Hand-pollination experiments were performed using five treatments: (i) agamospermous treatment—the pollinaria were removed before anthesis using forceps, and the flowers were then bagged (20 flowers from five individuals); (ii) autonomous autogamous treatment—flowers were bagged with a fine-meshed net before anthesis to exclude pollinators (20 flowers from five individuals); (iii) manually autogamous treatment—the pollinaria were removed and used to hand-pollinate the same flower before bagging (20 flowers from five individuals); (iv) manually allogamous treatment—same as treatment (iii) but using the pollinia from a different plant at least 1 m from the recipient plant (20 flowers from five individuals); and (v) open treatment—flowering individuals were randomly tagged and allowed to develop fruit under natural conditions (40 flowers from 10 individuals). The experimental plants were monitored intermittently over the subsequent 4–6 weeks; fruit set among the treatments was compared via Fisher’s exact test. Mature fruits were collected and silica-dried; seed mass was obtained to the nearest 0.0001 g. Thereafter, 200 seeds per capsule were examined to assess the presence of the embryo. After confirming the normality and homogeneity of variance using the Shapiro-Wilk and Bartlett’s tests, the effects of pollination treatment on the seed mass and the proportion of seeds with at least one embryo were tested via ANOVA.

Eleven G. crassifolia individuals representing six populations, ten G. schlechtendaliana individuals (including five of G. schlechtendaliana var. yakushimensis), and fifteen G. similis individuals were collected throughout Japan. Three individuals of G. × tamnaensis, a natural hybrid between G. schlechtendaliana and G. similis (Lee et al. 2010, 2012; Suetsugu et al. 2021b), were included in the comparative study (Suppl. material 1). Genomic DNA was extracted from silica-dried leaves using the CTAB method. An MIG-seq library for the 39 Goodyera samples was prepared according to the protocol outlined in Suyama et al. (2022). The library was sequenced using an Illumina MiSeq Sequencer (Illumina, San Diego, CA, USA) with a MiSeq Reagent Kit v3 (150 cycle, Illumina). The raw MIG-seq data of the 15 G. similis samples, 10 G. schlechtendaliana samples (including five G. schlechtendaliana var. yakushimensis samples), and three G. × tamnaensis samples had previously been deposited at the DDBJ Sequence Read Archive (DRA, accession number DRA011506) for Suetsugu et al. (2021b). The raw MIG-seq data of the 11 G. crassifolia samples were deposited at the DDBJ Sequence Read Archive (DRA, accession number DRA014540).

After removing the primer sequences and low-quality sequencing reads (Suetsugu et al. 2021b), 3 594 716 reads (92 172 ± 3937 reads per sample) were obtained from 4 058 158 raw reads (104 055 ± 4344 per sample). Stacks 2.60 pipeline was used for de novo single nucleotide polymorphism (SNP) discovery (Rochette et al. 2019), with the following parameters: minimum depth of coverage required to create a stack (m) = 3, maximum distance allowed between stacks (M) = 2, and number of mismatches allowed between sample loci while building the catalog (n) = 2. For the maximum likelihood and SplitsTree phylogenetic analyses, SNPs retained by four or more samples were used; for the population structure analysis, SNPs retained by 16 or more samples were used. SNPs with high heterozygosity (Ho ≥ 0.6) were removed. SNP sites with fewer than three minor alleles were filtered out. Finally, 4790 SNPs from 2795 loci were retained for phylogenetic analysis. For STRUCTURE analysis, to avoid linked SNPs, we used only the first SNP from each locus, retaining 874 SNPs.

Our SNP-based maximum likelihood phylogeny was inferred using RAxML 8.2.10 (Stamatakis 2014), using a GTR substitution model with Lewis’ ascertainment bias correction and 1000 iterations of parallelized tree search bootstrapping. To examine interspecific hybridization, a Neighbor-Net network was constructed using SplitsTree4 4.14 (Huson and Bryant 2006) using the uncorrelated P distance matrix. Population structure was examined using STRUCTURE 2.3.4 (Pritchard et al. 2000). We performed 20 independent runs, with a burn-in of 100 000 steps and an additional 100 000 steps using an admixture model, and estimated the log-likelihoods for each cluster (K = 1–10). Optimal K values were determined using the Delta K method (Evanno et al. 2005) in Structure Harvester (Earl and vonHoldt 2012). The results were visualized using CLUMPAK (Cluster Markov Packager Across K) (Kopelman et al. 2015).

The most remarkable characteristic of G. crassifolia is its column with lateral appendages (Figs 1–5). The lateral column appendages are consistently absent in the closely related taxa. Since the lateral appendages are themselves column-like, they are likely to be enlarged staminodes (Oh et al. 2022). Notably, the lateral appendages of the column differ significantly in size among populations, and in terms of their position on the inflorescence, being often conspicuous in the basal flowers and inconspicuous (or rarely absent) in the apical flowers. We observed an association between the column and lip or rostellum shape; the lip and the rostellum appeared to be three-lobed when the lateral appendages are conspicuous (Figs 2E, F, 3E, F, 4G, H, 5E, G). Given that the floral organ formation is explained mainly by the combined expression of ABCE-class MADS-box transcription factors (Causier et al. 2010; Hsu et al. 2015, 2021; Suetsugu et al. 2022), the spatial expression of the factors underlying this distinctive morphology deserves further investigation. In particular, the enlarged staminodes indicate that G. crassifolia exhibits some radial symmetry, unlike most orchid flowers, which are typically zygomorphic.

Figure 1. 

Goodyera crassifolia in its natural habitat A flowering individual B flowers C fruiting individual D leaves. Scale bars: 30 mm.

Figure 2. 

Goodyera crassifolia from Kami City, Kochi Prefecture (Hisanori Takeuchi G161-1, KYO) A dorsal sepal (abaxial view) B lateral sepals (left: abaxial view, right: adaxial view) C lateral petals (left: abaxial view, right: adaxial view) D lip and column (dorsal view) E lip (left: adaxial view, right: lateral view) F column (left: obliquely dorsal view, right: ventral view) G column (left: ventral view, right: lateral view) H lateral appendages removed from column (left: dorsal view, right: ventral view) I lateral appendages removed from column (both: dorsal view) J pollinarium (left: dorsal view, right: ventral view) K anther cap (left: dorsal view, right: ventral view). Arrows indicate the conspicuous lateral appendages. Photographs except G and I are derived from the same flower. G and I are used to show morphological variation of column within the same individual. Scale bars: 3 mm.

Figure 3. 

Goodyera crassifolia from Higashimuro County, Wakayama Prefecture (Yasuo Takada s.n., KYO) F, G column The conspicuous lateral appendages are indicated by arrows H column removing lateral appendages I, J lateral appendages removed from column K pollinarium L anther cap and pollinarium A dorsal sepal (abaxial view) B lateral sepals (left: abaxial view, right: adaxial view) C lateral petals (left: abaxial view, right: adaxial view) D lip and column (dorsal view) E lip (left: adaxial view, right: lateral view) F column (left: dorsal view, right: ventral view) G column (obliquely lateral view) H column removing lateral appendages (ventral view) I lateral appendages removed from column (left: dorsal view, right: ventral view J lateral appendages removed from column (ventral view) K pollinarium (ventral view) L anther cap and pollinarium (left: dorsal view, right: ventral view). Arrows indicate the conspicuous lateral appendages. Photographs except G, H, J, K are derived from the same flower G, H, J show the variation of column morphology within the same individual, while K is used because pollinaria were detached from anther cap of a flower that was mainly used. Scale bars: 3 mm.

Figure 4. 

Goodyera crassifolia (Koji Tanaka KS209, KYO; photographed after immersion in 50 percent ethanol) A flower (lateral view) B flower (dorsal view) C dorsal sepal (abaxial view) D lateral sepals (left: abaxial view, right: adaxial view) E lateral petals (left: abaxial view, right: adaxial view) F lip and column (lateral view) G lip (left: adaxial view, middle: lateral view, right: abaxial view) H column (left: dorsal view, right: obliquely ventral view) I column with partially detached lateral appendages (left: ventral view, right: lateral view) J lateral appendages removed from column (ventral view) K anther cap and pollinarium (ventral view). Arrows indicate the conspicuous lateral appendages. All photographs are derived from the same flower. Scale bars: 3 mm.

Figure 5. 

Goodyera crassifolia (Hisanori Takeuchi & Kenji Suetsugu KS208, KYO) A dorsal sepal (adaxial view) B lateral sepal (adaxial view) C lateral petal (adaxial view) D lip and column (dorsal view) E lip (left: adaxial view, right: lateral view) F longitudinal section of lip (adaxial view) G column and anther (left: top view, right: lateral view) H column (left: dorsal view, middle: lateral view, right: ventral view) I pollinarium (left: dorsal view, right: ventral view) J anther cap (dorsal view). Arrows indicate the conspicuous lateral appendages. All photographs are derived from the same flower. Scale bars: 3 mm.

Detailed morphological examination revealed that G. crassifolia can be distinguished from G. schlechtendaliana by not only column shape (column with vs. without lateral appendages) but also plant height (20–37 cm vs. ca. 15 cm), leaf texture (coriaceous vs. papyraceous), leaf coloration (glossy green, with narrow pale-white reticulation, to green with no decorations vs. green with obvious and broad white reticulation), inflorescence architecture (lax, internodes 17–24 mm long at inflorescence base vs. dense internodes 6–10 mm long at inflorescence base), pedicellate ovary length (11–20 mm, longer than floral bract vs. 7–9 mm, as long as the floral bract), flower opening (opening weakly vs. widely), flower size (sepal and petal length > 10 mm vs. < 10 mm), shape of lateral sepal (recurved at two-thirds of its entire length from the base vs. strongly recurved at half its entire length from the base), hypochile shape (weakly vs. strongly concave-saccate), and seed shape (often polyembryonic vs. always monoembryonic) (Lee et al. 2010, 2012; Bhattacharjee and Chowdhery 2012; Suetsugu and Hayakawa 2019; Suetsugu et al. 2021b; Oh et al. 2022).

It should be noted that G. crassifolia has previously been confused with G. × tamnaensis in Japan (Takahashi 1985; Akiyama 2010; The Flora-Kanagawa Association 2018). In fact, G. crassifolia is superficially similar to G. × tamnaensis in terms of its weakly opening flowers but differs in plant height (20–37 cm for G. crassifolia vs. 10–15 cm for G. × tamnaensis), leaf texture (coriaceous vs. papyraceous), leaf coloration (glossy green with narrow, pale-white reticulation to green with no decoration on upper surface vs. velutinous dark green with a white central vein and reticulate venation), ovary and pedicel length (11–20 mm vs. 7–10 mm long), flower size (petal and sepal length > 10 mm vs. < 10 mm), column shape (column with vs. without lateral appendages), and rostellum shape (acuminate apex, occasionally bi- or trilobed vs. flattened and cuneate apex, never divided) (Lee et al. 2010, 2012; Bhattacharjee and Chowdhery 2012; Suetsugu and Hayakawa 2019; Suetsugu et al. 2021b).

Further detailed comparison of morphological characters among G. crassifolia, G. schlechtendaliana and G. × tamnaensis is given in Table 1. Additional descriptions and illustrations of G. crassifolia, G. schlechtendaliana, G. × tamnaensis, and G. similis are available in Lee et al. (2010, 2012), Suetsugu and Hayakawa (2019), Suetsugu et al. (2021b), and Oh et al. (2022).

Polyploidization is commonly accepted as a vital mechanism of sympatric speciation in plants (Köhler et al. 2010). Owing to chromosome number imbalance during meiosis, backcross between either parent would mostly result in nonviable progenies; those rare survivors with unbalanced chromosome numbers will be primarily sterile (Ramsey and Schemske 1998). The triploid-block is a significant reproductive barrier leading to polyploid speciation (Köhler et al. 2010).

Investigation of chromosome numbers provided evidence of polyploidy in G. crassifolia: all of the G. schlechtendaliana individuals (including G. schlechtendaliana var. yakushimensis) showed a chromosome number of 2n = 30; whereas all G. crassifolia individuals (Fig. 6) showed 2n = 60. In line with the results obtained in this study, Oh et al. (2022) reported 2n = 60 for a Korean G. crassifolia individual. Intriguingly, Sera (1990) reported 2n = 60 in five “G. schlechtendaliana” plants from four localities, while reporting 2n = 30 for most G. schlechtendaliana individuals (60 plants from 22 localities collected throughout Japan). However, Sera (1990) noted that the 2n = 60 “G. schlechtendaliana” plants possess coriaceous leaves with faint reticulate variegation. The photographs listed in Sera (1990) indicate that they also have laxly flowered inflorescences and a longer pedicellate ovary, which are characteristic features of G. crassifolia. Although three of the voucher specimens from Sera (1990) have unfortunately been lost, possibly during the relocation of the herbarium HIBG (T. Sera, personal communication), we could identify the two remaining voucher specimens as G. crassifolia. It is likely that all of the 2n = 60 plants of Sera (1990) could be G. crassifolia. Given that 2n = 30 is the only chromosome number reported in G. schlechtendaliana as determined by other previous studies (Matsuura and Nakahira 1958; Shoji 1963; Tanaka 1965; Sun et al. 1996; Tae et al. 1997), 2n = 30 is arguably the typical chromosome number of G. schlechtendaliana. In addition, 2n = 30 has been reported in G. × tamnaensis (Lee et al. 2012), although speciation via hybridization without a change in chromosome number is considered rare (Schumer et al. 2014). Thus, as suggested by Oh et al. (2022), the cytological distinctness of G. crassifolia may have partially contributed to its reproductive isolation.

Figure 6. 

Somatic chromosomes (A–C) and their explanatory drawings (D–F) of Goodyera crassifolia and its closely related taxa A, D G. crassifolia B, E G. schlechtendaliana C, F G. schlechtendaliana var. yakushimensis. Scale bars: 10 μm.

Our pollination experiments revealed the contrasting breeding systems of G. crassifolia and G. schlechtendaliana. The latter, although self-compatible, is neither autogamous nor agamospermous, and shows low fruit set under natural conditions; pollinator limitation was the major cause of low fruit set, which was significantly improved by manual autogamy and allogamy (Table 2, P < 0.001). By contrast, the natural populations of G. crassifolia consistently exhibited high fruit set (Fig. 1). Given that high fruit set was obtained in agamospermous, bagged, manually geitonogamous, manually allogamous, and open flowers, G. crassifolia flowers are not pollinator-limited under natural conditions. Neither seed mass nor the proportion of seeds with embryo varied significantly with pollination treatment (Table 2). Given that the rostellum functionally prevents autonomous autogamy, agamospermy is arguably the main cause of high fruit set in G. crassifolia. Therefore, agamospermy provides reproductive assurance under pollinator limitation in G. crassifolia.

Notably, the viscidium of G. crassifolia exhibits almost no adhesion, hindering its attachment onto its potential pollinators. No pollinia removal or deposition was observed during the field study. Because (i) G. crassifolia has weakly opened flowers with less-adhesive pollinia and (ii) its stigma is sometimes covered with column appendages (Figs 2G, 3F), there are arguably few opportunities for outcrossing. Thus, agamospermy is probably its dominant, if not exclusive, reproductive strategy. The reduced selection pressure on outcrossing may have led to the aforementioned variations in the lip, column appendages, and rostellum morphology, even within a single inflorescence. The polyembryony detected in G. crassifolia is further indicative of agamospermy, given that adventitious embryony, the most common form of apomixis, is characterized by a high number of polyembryonic seeds (Catling 1982; Campacci et al. 2017; Naumova 2018).

During our field study, we confirmed the phenological isolation between G. crassifolia and G. schlechtendaliana as previously reported by Takahashi (1985) and Oh et al. (2022). In many regions where both are sympatric (e.g., Hongdo, Korea: Oh et al. 2022; southern and central Japan: Takahashi (1985) and field observations in this study), G. schlechtendaliana starts to flower ca. 3–4 weeks earlier than G. crassifolia. Despite the slight overlap in their flowering periods, the temporal isolation could significantly reduce interspecific cross-pollination. In addition, the predominantly agamospermous breeding system of G. crassifolia helps maintain its reproductive isolation from G. schlechtendaliana. A similar reproductive isolation mechanism was proposed to explain the maintenance of integrity between sexually reproducing taxa and agamospermous taxa within the same genus (Catling and Brown 1983).

MIG-seq-based maximum likelihood phylogenetic tree generated in this study revealed that G. crassifolia forms a separate clade from G. similis and G. schlechtendaliana (100% bootstrap value: Fig. 7). Goodyera schlechtendaliana was paraphyletic, while the monophyly of G. schlechtendaliana var. yakushimensis was supported (100% bootstrap value). Neighbor-Net phylogenetic analysis indicated that G. crassifolia, G. schlechtendaliana, and G. similis represent three distinct genetic clusters (Fig. 8). In the Neighbor-Net analysis, we show that the genetic diversity of G. schlechtendaliana as a whole, including G. schlechtendaliana var. yakushimensis, is comparable to that of G. similis. Therefore, G. schlechtendaliana var. yakushimensis is more likely to be an intraspecific variant of G. schlechtendaliana rather than an independent species. The interpretation is also based on the results of STRUCTURE analysis mentioned below, as well as on the relatively small morphological differences between var. schlechtendaliana and var. yakushimensis indicated by Suetsugu and Hayakawa (2019).

Figure 7. 

Phylogenetic tree of Goodyera crassifolia and its closely related taxa reconstructed using MIG-seq data. Bootstrap values within species, and those less than 50%, are not shown. Branch length represents the average number of substitutions per site.

Figure 8. 

Neighbor-Net network for Goodyera crassifolia and its closely related taxa, based on uncorrected P distances calculated from 4790 SNPs.

The STRUCTURE analysis at K = 2 (the largest delta K for our data) classified G. crassifolia and G. schlechtendaliana (including var. yakushimensis) into the same cluster, while at K = 3 (the second-largest delta K), G. crassifolia, G. schlechtendaliana (including var. yakushimensis), and G. similis formed three groups (Fig. 9). These findings, together with its multiple morphological differences from those observed in G. schlechtendaliana and G. similis, support the status of G. crassifolia as an independent species. Furthermore, genetic variation, which was high in the outcrossing G. schlechtendaliana, was low in the predominantly agamospermous G. crassifolia, both between and within populations. Similar patterns have been observed in other orchids, including Nigritella Rich., which includes both outcrossing and agamospermous species (Hedrén et al. 2018).

Figure 9. 

Population structure of Goodyera crassifolia and its closely related taxa, inferred with STRUCTURE 2.3.4. Using K = 2 and K = 3 generated the largest and second-largest delta K, indicating that they were the most and second most optimal, respectively. Species and populations are separated by broad and narrow vertical black lines, respectively.

Molecular data obtained in this study provide further evidence that G. crassifolia has a different evolutionary origin from G. × tamnaensis. Both phylogenetic and population structure analyses showed that G. × tamnaensis has genetic components of both G. schlechtendaliana and G. similis (Figs 7–9). By contrast, although G. crassifolia was suspected as a natural hybrid of G. schlechtendaliana and G. similis (Takahashi 1985; Akiyama 2010), neither the phylogenetic analysis nor the population structure analyses support genetic admixture between G. crassifolia and any of its close congeners (Figs 7–9).