Total DNA was extracted using a modified CTAB-Qiagen column protocol (Kuo 2015). Three plastid DNA regions, rbcL, rps4-trnS (rps4 gene + rps4-trnS intergenic spacer), and trnL-F (trnL gene + trnL-trnF intergenic spacer), were amplified and sequenced using the primers “ESRBCL1F” and “1379R” for rbcL (Pryer et al. 2001, Schuettpelz and Pryer 2007), “RPS5F” and “TRNSR” for rps4-trnS (Nadot et al. 1995, Smith and Cranfill 2002), and “FernL 1Ir1” and “f” for trnL-F (Taberlet et al. 1991, Li et al. 2010).

The PCR amplifications were performed in 16 μl reactions containing ca. 10 ng template DNA, 1×Taq DNA Polymerase Master Mix RED solution (Ampliqon, Denmark), and 1 μl each of 10 μM primers. The PCR reactions were carried out in a GeneAmp PCR System 9700 (Applied Biosystems, Carlsbad, California, USA). Thermocycling conditions were the same for PCRs of these three regions and comprised an initial denaturation of 2 minutes at 94°C followed by a core sequence of 35 repetitions of 94°C for 1 minute, 55°C for 1 minute, and 72°C for 1 minute followed by a final extension of 10 minutes at 72°C. Resulting PCR products were sequenced using the same PCR primers with BigDye^TM terminator (Applied Biosystems, Carlsbad, California, USA). The newly generated sequences were deposited in GenBank. GenBank accession numbers and voucher information are provided in Appendix.

Initial BLAST against the NCBI nucleotide database (Altschul et al. 1990) based on rbcL sequences indicated that Dryopolystichum phaeostigma is closely related to the species of Polypodiineae families, including Lomariopsidaceae, Nephrolepidaceae, Tectariaceae, and Dryopteridaceae. Accordingly, we assembled a data matrix including 250 species representing 36 genera from these families (Appendix). Sampling included all the four genera in which D. phaeostigma has been placed (i.e., Dryopteris, Polystichum, and Tectaria).

Sequences were aligned using Geneious v6.1.8 (Drummond et al. 2011) and then manually checked for errors. The three single-region (rbcL, rps4-trnS, and trnL-F) and dataset combining all three were independently subjected to both maximum likelihood (ML) and Bayesian inference (BI) phylogenetic analyses. Data matrices are available in TreeBASE, study number 20506, at https://treebase.org/. ML tree searches were conducted using RAxML (Stamatakis 2006) employing the GTRGAMMA substitution model through the CIPRES portal (Miller et al. 2010). Five independent searches for the ‘best tree’ and 1,000 bootstrap replicates were performed using a region-partitioned dataset. BI analyses were conducted using MrBayes 3.2.1 (Ronquist and Huelsenbeck, 2003) employing the same substitution model as in ML analysis. Each analysis consisted of two independent runs with four chains for 10⁶ generations, sampling one tree every 1000 generations. Burn-in was set to 10000 based on our preliminary analysis. The convergences of MCMC runs were checked using Tracer v.1.6 (Rambaut et al. 2014).

We addressed the possibility of phylogenetic bias due to long branches following the recommendation of Siddal and Whiting (1999). Since Dracoglossum and Lomariopsis were resolved on long branches in preliminary analyses (not shown), we conducted two additional analyses in which each one of the two long-branched genera, Dracoglossum and Lomariopsis, was excluded to examine whether phylogenetic placement and branch support for Dryopolystichum’s placement changed. Since maximum parsimony (MP) phylogeny is considered to be more susceptible to long-branch attraction (Philippe et al. 2005), we analyzed the concatenated dataset under MP in order to compare those results with our ML phylogeny. The MP analyses were conducted using TNT (Goloboff et al. 2008) following the search strategy detailed in Sundue et al. (2014).

All single-region phylogenies resolved Dryopolystichum phaeostigma in Lomariopsidaceae, but with two slightly different topologies. The rbcL and rps4-trnS phylogenies placed D. phaeostigma sister to a clade of Dracoglossum + Lomariopsis with 93% and 72% maximum likelihood bootstrap percentages (BS), respectively (Suppl. materials 2, 3). In comparison, the trnL-F phylogeny placed D. phaeostigma sister to Cyclopeltis (BS = 74%), and Dryopolystichum + Cyclopeltis was sister to Dracoglossum + Lomariopsis (Suppl. material 4). There was no strongly supported conflict between the ML and BI phylogenies (Suppl. materials 1–4). Both the ML and BI phylogenies based on the combined dataset (Fig. 2, Suppl. material 1) reveal the same topology as those based on the rbcL and rps4-trnS regions. Bootstrap support and posteriori probability (PP) for the above relationships were generally very high except for the branches placing D. phaeostigma, where BS was ≤ 70% and PP were ≤ 0.9 in all the phylogenies.

Figure 2. 

Simplified maximum likelihood phylogram of Polypodiineae obtained from the rbcL + rps4-trnS + trnL-F combined dataset. Maximum likelihood bootstrap percentages (BS) are provided at each node. Thickened lines indicate Bayesian inference posterior probability (PP) ≥ 0.9. Original phylogram with support values for all the nodes is available in Suppl. materials 1. Voucher information and GenBank accession numbers are shown in Appendix.

Removing Dracoglossum from the analysis had little effect on the topology within Lomariopsidaceae, and BS supports for the generic placement of Dryopolystichum remained low (≤ 70%, data not shown). In contrast, the removal of Lomariopsis resulted in higher BS values for all clades within Lomariopsidaceae (≥ 99%, data not shown). MP analyses also resulted in a clade comprising all the Lomariopsidaceae genera and Dryopolystichum, but Dryopolystichum was resolved as sister to Cyclopeltis (data not shown).

All examined sporangia (SITW10443) produced 32 normal spores, and the mean spore length was 64.1 ± 4.5 μm (Fig. 3). The chromosome number of the three sporophyte somatic cells observed was ca. 164 (Fig. 4). Results of flow cytometry revealed that the genome size of spore nuclei is approximately half of those of leaf nuclei (Fig. 5).

The reconstructed maximum likelihood and Bayesian inference phylogenies unambiguously resolved Dryopolystichum within Lomariopsidaceae (Fig. 2), a position not previously suggested (Kramer and Green 1990, Smith et al. 2006, Christenhusz et al. 2011, PPG I 2016). This placement is consistent in all our analyses. Nonetheless, the generic position of Dryopolystichum within Lomariopsidaceae remains poorly resolved. This uncertainty may be partially explained by the incongruence between trnL-F and the other analyzed regions, but our process of removing the long-branched genera showed that low BS was retrieved only when Dryopolystichum and Lomariopsis were both included in the analysis. These results may also be explained by the large amounts of missing data in Lomariopsis; 19 of the 25 species included were represented by trnL-F data alone. We recommend further phylogenetic study using an expanded dataset to resolve the intergeneric relationships within Lomariopsidaceae.

Phylogenetic analyses using DNA sequences have served as the basis for redrawing fern classifications in the 21^th century (Smith et al. 2006, Christenhusz et al. 2011, PPG I 2016). With respect to family circumscription, one of the most dramatically changed families is Lomariopsidaceae (Tsutsumi and Kato 2006, Schuettpelz and Pryer 2007, Christenhusz et al. 2013). Just prior to the molecular era, Lomariopsidaceae was treated as one of the largest fern families with six genera and over 500 species (e.g., Kramer and Green 1990) and was strongly supported by the following combination of characters: rhizomes with ventral root insertion, dictyosteles with elongate ventral meristeles, and dimorphic leaves where the fertile leaves had acrostichoid sori (Holttum and Hennipman 1959, Kramer and Green 1990).

Subsequent molecular phylogenetic analyses demonstrated that most genera previously treated in Lomariopsidaceae should be transferred to Dryopteridaceae (Tsutsumi and Kato 2006, Schuettpelz and Pryer 2007). The combination of characters uniting the former Lomariopsidaceae are now interpreted to have evolved multiple times, and to be correlated with dorsiventrality of the rhizome (Moran et al. 2010, McKeown et al. 2012). Meanwhile, Cyclopeltis was transferred from Dryopteridaceae to Lomariopsidaceae as suggested by molecular phylogeny (Schuettpelz and Pryer 2007), although it has none of the characters formerly used to circumscribe Lomariopsidaceae (Holttum and Hennipman 1959, Kramer and Green 1990).

More recently, the neotropical genus Dracoglossum was established (Christenhusz 2007) and later transferred to Lomariopsidaceae from Tectariaceae based on a molecular phylogeny (Christenhusz et al. 2013). This pattern was also unexpected since there are essentially no shared morphological characters by Dracoglossum and Lomariopsis, except for the ribbon-like gametophyte (R. C. Moran pers. com.). Our finding, that Dryopolystichum belongs to Lomariopsidaceae, comes as a further surprise. With these changes, Lomariopsidaceae is a family of five genera (Cyclopeltis, Dracoglossum, Dryopolystichum, Lomariopsis, and Thysanosoria) and ca. 70 species. As far as we can tell, none of the morphological traits commonly used unify these genera (Table 1). In the following paragraphs, we provide a recircumscription of both Lomariopsidaceae and Dryopolystichum, and then discuss selected characters in the light of our phylogenetic placement.

Perine architecture of Dryopolystichum is very similar to that of Dracoglossum plantagineum (Christenhusz 2007, Fig. 3). They are loosely attached, forming thin crests, and having a spiculate microstructure. Perine of Cyclopeltis and Thysanosoria are also similar in being loosely attached and having a spiculate microstructure, but they differ by having broader folds (Holttum and Hennipman 1959, Tryon and Lugardon 1991). The perine characters, however, are not shared by all the taxa of Lomariopsidaceae especially considering the variation of ornamentation existing in Lomariopsis (Rouhan et al. 2007). Moreover, these perine characters also appear in other Polypodiineae lineages particularly in bolbitidoid ferns (Moran et al. 2010) as well as in various Aspleniineae lineages (Sundue and Rothfels 2014, PPG I 2016).

Figure 3. 

Spores SEM of Dryopolystichum phaeostigma. A Lateral view of the spore B Detail of surface. Scale bars: A = 50 μm, B = 10 μm.

Blackish sclerenchyma strands are visible in the rhizome sections of Dryopolystichum (Fig. 1F). These are also present in Dracoglossum, Cyclopeltis, and Lomariopsis, but similar characters are known from various groups throughout Polypodiineae (Hennipman 1977, Moran 1986, Hovenkamp 1998). Further studies might reveal variation in these strands to be of systematic value.

The rachis-costae architecture of Dryopolystichum is characterized by an adaxially sulcate rachis with grooves that do not connect to those of the pinna-costae. The rachis is also narrowly winged laterally. Both characters are seen in Thysanosoria and in some species of Lomariopsis (Holttum and Hennipman 1959, Moran 2000). In contrast, Dracoglossum and Cyclopeltis have non-winged and non-sulcate rachises (Holttum 1991, Christenhusz 2007).

The chromosome number in somatic cells of Dryopolystichum phaeostigma was ca. 164 (Fig. 4). The base numbers for Lomariopsidaceae genera (Cyclopeltis, Dracoglossum, and Lomariopsis) are 40 or 41 (Walker 1985, Kato and Nakato 1999, Moran 2000), suggesting that D. phaeostigma is a tetraploid.

Figure 4. 

Chromosome number of Dryopolystichum phaeostigma. A Chromosomes at mitosis metaphase, 2n = ca. 164 (SITW10443) B explanatory illustration of A. Scale bars = 10 μm.

Our flow cytometry and spore count results indicate that Dryopolystichum phaeostigma is sexually reproducing and has 32 spores per sporangium (Fig. 5). In Polypodiales, sporogenesis leading to the formation of 64 spores in a sporangium is by far the most common pattern of sexually reproducing species, e.g., Aspleniaceae (Gabancho et al. 2010), Athyriaceae (Kato et al. 1992, Takamiya et al. 1999), Davalliaceae (Chen et al. 2014), Dryopteridaceae (Lu et al. 2006), Polypodiaceae (Wang et al. 2011), Pteridaceae (Huang et al. 2006), and Thelypteridaceae (Ebihara et al. 2014). Cases of sporogenesis resulting in 32 spores per sporangium are known from a few Polypodiales ferns but all belong to the suborders Lindsaeineae and Pteridineae, i.e., Lindsaeaceae (Lin et al. 1990), Cystodiaceae (Gastony 1981), and Ceratopteris (Pteridaceae; Lloyd 1973). Our study provides the first confirmed case of a sexual reproduction with 32 spores per sporangium in the suborder Polypodiineae.

Figure 5. 

Relative DNA contents of Dryopolystichum phaeostigma spore and leaf nuclei inferred by flow cytometry.