Phylogeny and Taxonomy on Cryptic Species of Forked Ferns of Asia

Cryptic species comprise two or more taxa that are grounded under a single name because they are more-or-less indistinguishable morphologically. These species are potentially important for detailed assessments of biodiversity, but there now appear to be many more cryptic species than previously estimated. One taxonomic group likely to contain many cryptic species is Dicranopteris, a genus of forked ferns that occurs commonly along roadsides in Asia. The genus has a complex taxonomical history, and D. linearis has been particularly challenging with many intra-specific taxa dubiously erected to accommodate morphological variation that lacks clear discontinuities. To resolve species boundaries within Dicranopteris, we applied a molecular phylogenetic approach as complementary to morphology. Specifically, we used five chloroplast gene regions (rbcL, atpB, rps4, matK, and trnL-trnF) to generate a well-resolved phylogeny based on 37 samples representing 13 taxa of Dicranopteris, spanning the major distributional area in Asia. The results showed that Dicranopteris consists of ten highly supported clades, and D. linearis is polyphyletic, suggesting cryptic diversity within the species. Further through morphological comparison, we certainly erected Dicranopteris austrosinensis Y.H. Yan & Z.Y. Wei sp. nov. and Dicranopteris baliensis Y.H. Yan & Z.Y. Wei sp. nov. as distinct species and proposed five new combinations. We also inferred that the extant diversity of the genus Dicranopteris may result from relatively recent diversification in the Miocene based on divergence time dating. Overall, our study not only provided additional insights on the Gleicheniaceae tree of life, but also served as a case of integrating molecular and morphological approaches to elucidate cryptic diversity in taxonomically difficult groups.


INTRODUCTION
Cryptic species, a common and increasingly used term, refers to taxa that are erroneously classified as a single species due to the paucity of conspicuous morphological differences (Trontelj and Fišer, 2009;Detwiler et al., 2010;Struck et al., 2018). Cryptic species represent a potentially important influence on the accuracy of detailed assessments of biodiversity (Witt et al., 2006;Trontelj and Fišer, 2009) and can lead to novel insights regarding patterns and processes of biodiversity, including geographic variation in species distributions and species coexistence (Fiser et al., 2018). However, cryptic species are seldom considered in biodiversity assessments owing to the lack of affordable and efficient diagnostic methods (Witt et al., 2006). This is compounded by the fact the high rate at which cryptic species are discovered in molecular studies suggests that number of cryptic species is far greater than prior estimates.
Studies on cryptic species throughout the whole tree of life have increased exponentially over the past two decades, fueled in large part by the increasing availability of DNA sequences, which facilitate various genetic approaches to the resolution of cryptic diversity (Sites and Marshall, 2003;Bickford et al., 2007). The prevalence of considerable cryptic diversity has been uncovered in a diverse range of groups, including in plants (Okuyama and Kato, 2009;Carstens and Satler, 2013;Ji et al., 2020;Kinosian et al., 2020;Li et al., 2020) and many animals (Hebert et al., 2004;Oliver et al., 2009;Manthey et al., 2011;Nadler and DE León, 2011;Phiri and Daniels, 2016), suggesting that cryptic species probably represent a significant portion of undiscovered biodiversity (Jörger and Schrödl, 2013;Pante et al., 2015;Loxdale et al., 2016). The ever-increasing cryptic diversity that genetics has resolved poses a taxonomic challenge in terms of what taxonomic ranks should be assigned to cryptic species that can be recognized on a genetic, but not necessarily morphological, basis. One practical proposal has been that the taxonomic ranks of species should be reserved for organisms showing observable morphological variation, while resolved cryptic species should be designated as intraspecific ranks (Maxwell and Dekkers, 2019;Maxwell et al., 2021).
Cryptic species may also be concealed in cases where there is considerable morphological variation but without clear boundaries supporting species delimitation, such as in the fern genus Dicranopteris Bernh. Dicranopteris is one of six genera in Gleicheniaceae (PPG I, 2016), an early-diverging leptosporangiate fern (Schuettpelz and Pryer, 2007;Choo and Escapa, 2018). The genus is unique on the part of its branching and leaf morphology as characterized by a pseudo-dichotomously branched fronds that produces a forking architecture, resulting from abortion or dormancy of the apical bud (Perrie et al., 2007). The unique morphology is attractive by which some species were used as ornamental plants (Marpaung and Susandarini, 2021). Dicranopteris is composed of about 20 species (Schuettpelz et al., 2016) that occur widely in tropical and subtropical areas (Ding et al., 2013). Across its range, Dicranopteris displays an extraordinarily high level of morphological diversity (Figure 1) that is not easily translated into species boundaries. In fact, due to these difficulties, no comprehensive species classification has been proposed for the genus, despite several prior studies based on morphology, sometimes with only regional sampling (Supplementary Table 1).
Within Dicranopteris, D. linearis (Burm.f.) Underw. has been particularly taxonomically challenging. D. linearis was divided into 13 varieties in Southeast Asia by Holttum (1959), who admitted that some varieties were more distinct than others and should probably be recognized as species. For example, the four varieties D. linearis var. demota, D. linearis var. tetraphylla, D. linearis var. montana, and D. linearis var. altissima, always possess accessory branches at the bases of their ultimate branches, whereas other varieties are not always present. In addition, these four varieties can be also separated based on other characteristics, such as branches at successive forks alternately being equal or unequal, lower pinnae surfaces bearing some hairs or not, angles of secondary and later forks, and dimensions of the pinnae (Underwood, 1907;Nakai, 1950;Ching et al., 1959;Holttum, 1959;Huang, 1994;Piggott, 1996;Ding et al., 2013). Following Holttum, Ding et al. (2013) recognized D. linearis and its varieties as a single species and placed them under the D. pedata, which likely obscured its real complexity. With these regards, taxonomic revisions on the placement of D. linearis and its varieties into the genus are very dynamic. Nonetheless, several varieties have been reinstated the status of species. A study by Ding et al. (2013) on the genus treated D. linearis var. montana under D. taiwanensis. Similarly, Kuo et al. (2019) in their phylogenetic study of updating Taiwanese pteridophyte checklist provided support for separating Dicranopteris subpectinata (Christ) C.M. Kuo and Dicranopteris tetraphylla (Rosenst.) C.M. Kuo of Taiwan from D. linearis.
Due to the diverse morphology and complex taxonomic history of Dicranopteris (Supplementary Table 1), we suspected that there may be a large number of cryptic species in this genus. Therefore, we sought to apply molecular phylogeny as a complement to morphology to resolve its species boundaries. To accomplish this, we incorporated 13 taxa, representing well over half of the known diversity in Dicranopteris, and five plastid DNA regions (rbcL, atpB, matK, rps4, and trnL-trnF). Specifically, our aims were to: (1) resolve phylogenetic relationships in Dicranopteris, especially among the varieties of D. linearis, (2) provide a preliminary revised taxonomy of Dicranopteris on the basis of molecular phylogeny and morphological characters, and (3) uncover cryptic diversity in the genus if present.

Taxon Sampling
We sampled 37 accessions representing 13 taxa of Dicranopteris, spanning the major centers of geographic distribution in Asia (Supplementary Table 2): China (23 accessions), Malaysia (7 accessions), Thailand (3 accessions), and Indonesia (4 accessions). Our sampling followed taxonomic treatments in Holttum (1959); Ding et al. (2013), and Schuettpelz et al. (2016). We also sampled close relatives for the outgroup: Dipteris chinensis, Matonia pectinata, Sticherus truncatus, and Diplopterygium glaucum. Notably, all samples were approved by for collection in China or provided by the collaborators in Malaysia, Indonesia, or Thailand. We deposited duplicates of all voucher specimens at the Shanghai Chenshan Herbarium (CSH) and the National Orchid Conservation Center of China and the Orchid Conservation and Research Center of Shenzhen (NOCC).

DNA Extraction and Sequencing
We first extracted total genomic DNA from silica-gel dried leaves using a Plant Genomic DNA Kit (Tiangen Biotech, Beijing, China) following the manufacturer's protocol. From the total DNA, we amplified five chloroplast DNA (cpDNA) regions including four coding regions (rbcL, atpB, matK, rps4) and one intergenic spacer (trnL-trnF). We performed polymerase chain reaction (PCR) amplification and sequencing for these gene regions using the primers shown in Table 1. The amplification procedure for all regions consisted of a 25-µl reaction volume containing 1-2 µl of template DNA, 1 µl each of 10 µM primers, 2.5 µl of 10× Taq Buffer with MgCl 2 , 0.2 µl of Taq polymerase, and ddH 2 O to volume. The amplification profiles comprised initial enzyme activation at 95 • C for 5 min followed by 38 cycles of denaturation, primer binding, and extension at 94 • C for 30 s, 58 • C for 30 s, and 72 • C for 1 min, respectively. The final extension was at 72 • C for 10 min. The resulting PCR products were purified and sequenced in Sangon Biotech. All DNA sequences new to this study are available from GenBank (accession numbers in Supplementary Table 2).

Phylogenetic Analyses
We edited and assembled the resulting sequences using SeqMan v7.1.0 (DNASTAR, United States) and aligned them using MUSCLE (Edgar, 2004) with default parameters followed by manual adjustment. Then, the alignments of the five cpDNA regions were concatenated into a single combined dataset using PhyloSuite (Zhang et al., 2020). Prior to performing phylogenetic analyses, we selected optimal partitioning strategies and best substitution models using PartitionFinder 2 (Lanfear et al., 2017) integrated into PhyloSuite, with three codon positions for each protein-coding gene provided as input partitions. The optimized partitioning schemes, and associated models are provided in Supplementary Table 1. We further utilized these optimized partitioning schemes and models to perform Bayesian inference (BI) and maximum likelihood (ML) analysis, respectively. Maximum likelihood phylogeny was inferred using IQ-TREE (Nguyen et al., 2015) integrated into PhyloSuite under each edgelinked partition model with 5,000 ultrafast bootstrap replicates (Minh et al., 2013). For Bayesian inference, we performed the analyses in MrBayes v3.2.6 (Ronquist et al., 2012) with the optimized partitioning schemes and models. The BI analysis comprised two parallel runs of four Markov chain Monte Carlo (MCMC) each for 1,000,000 generations with every 1,000 generations sampled. The standard deviation of split frequencies was set to less than 0.01 to achieve the convergence of the independent runs. Following the analysis, we removed the first 25% of sampled generations as burn-in. Additionally, we carried out a maximum parsimony (MP) analysis on the concatenated dataset using MPBoot (Hoang et al., 2018) with 1,000 bootstrap replicates (-bb 1,000).

Molecular Dating Analysis
The divergence time estimations were conducted using the concatenated cpDNA dataset in BEAST2 v2.6.3 (Bouckaert et al., 2019). We calibrated the BEAST analysis based on two fossils following Schuettpelz and Pryer (2009) suggesting ≥228 million years ago (Mya) as the divergence between Matoniaceae and Dipteridaceae along with ≥99.6 Mya for divergence of the clade consisting of Sticherus and Diplopterygium. We also used the results from Schuettpelz and Pryer (2009) to constrain the root of Gleicheniales (= 262.2 Mya) as a secondary calibration point with a normal distribution.
The Bayesian Evolutionary Analysis Utility (BEAUTi) (BEAST package) was utilized to generate an XML file for the analysis in BEAST, in which we applied a GTR model of nucleotide substitution with four gamma rate categories and an uncorrelated lognormally distributed relaxed (UCLD) model for the molecular clock. We ran the MCMC chain for 30 million generations with 25% burn-in and a sampling frequency of 30,000 generations. The tree branching process was inferred using the Yule model with other default settings. We used Tracer v1.7.1 (Rambaut et al., 2018) to check the effective sample size (ESS) of each parameter and found that all ESS exceeded 200, which is considered as recommended threshold to indicate stationarity (Drummond et al., 2012). We determined the maximum clade credibility tree using TreeAnnotator v2.6.3 (BEAST package) with median node heights and visualized it in FigTree v1.4.3 (Rambaut, 2016).

Hidden Cryptic Biodiversity in Dicranopteris
Evolutionary stasis is common for ancient lineages (Bomfleur et al., 2014;Jin et al., 2020;Wei et al., 2021), such as ferns. Accordingly, cryptic species may be particularly prevalent among these lineages due to extensive phenotypic plasticity, often resulting in extreme difficulty in classifying old taxa . In this study, we found that the main extant lineages of Dicranopteris were ancient (Figure 3), which is consistent with the cryptic diversity in the genus that we detected as well as ongoing taxonomic difficulties. Compounding the problem is that Dicranopteris may have experienced flourishing diversification during the Paleocene to the Micoene (Schneider et al., 2004;Schuettpelz and Pryer, 2007;Sundue et al., 2015; Figures 2, 3 and Supplementary Figure 1), and this can also lead to subsequent taxonomic confusion.
Recently, Marpaung and Susandarini (2021) investigated the variation on morphology and spore characters among two species and seven varieties of Dicranopteris. The result revealed that the genus has high morphological variations, as indicated by the presence of intraspecific categories. In the present study, we further used molecular data as complementary to morphology to explore cryptic biodiversity in Dicranopteris. We were able to phylogenetically and morphologically differentiate two new species (Figures 2, 4), D. austrosinensis sp. nov. and D. baliensis sp. nov., and resolve several previously unsettled  aspects of species classification of Dicranopteris, including five new combinations. In addition, the species D. pedata recorded in Flora of China (Ding et al., 2013) was proved to hide cryptic species. Specifically, the current treatment of D. pedata in Flora of China consists of three species, D. linearis, D. austrosinensis sp. nov., and D. pedata s. str. Among the three species, only D. austrosinensis sp. nov. presents a pair of accessory branches at the ultimate branches. D. linearis differs from the similar D. pedata by having more basal inner segments of ultimate pinnules shortened. Overall, we have preliminarily revealed that Dicranopteris likely harbors a large constellation of cryptic diversity (Figure 2 and Supplementary  Figures 1, 2).

Preliminary Phylogenetic Analyses
The molecular phylogenetic relationships of Dicranopteris have never been comprehensively investigated in prior studies. Previous studies have included only several well-accepted species of the genus, such as D. pedata and D. linearis (Hasebe et al., 1995;Schuettpelz and Pryer, 2007), within studies of broader taxonomic groups. In this study, we included accessions representing well over half of the currently recognized diversity of Dicranopteris and utilized three methods of phylogenetic reconstruction (ML, BI, and MP).
The preliminary phylogenetic analyses showed a clear division of species in Dicranopteris, which was generally congruent with previous study based on morphology (Ching et al., 1959), except for varieties of D. linearis. D. linearis exhibited no close relationship with its varieties, while was typically well supported as monophyletic with D. taiwanensis (BIPP = 1.0, MLBS = 86%, and MPBS = 88%) (Figure 2 and Supplementary Figures 1, 2). In addition, D. linearis was proved to be a distinct species and not a synonym for D. pedata as previously thought (Ding et al., 2013; Figure 2 and Supplementary Figures 1, 2). For the four species (i.e., D. alternans, D. inaequalis, D. subspeciosa, and D. subpectinata), the species status and phylogenetic relationship were not well-documented and untangled up in the present study (Figure 2 and Supplementary Figures 1, 2) due to in the absence of adequate samples. Conversely, they can be identified by many morphological characters, such as angle of the forks, lower surface covering, etc. Nonetheless, we believe that classifying them would be premature, as we seek stable taxonomic solutions within the genus and for the D. linearis complex.
In a broad sense, the phylogenetic positions of several main clades of Dicranopteris had low support (i.e., BIPP < 0.90, MLBS < 70%, and MPBS < 70%). Consequently, in the future, more molecular markers and denser sampling of taxa should be included to better resolve taxonomic relationships. Also, a more comprehensive systematic investigation of Dicranopteris should accompany the ongoing work on the Flora of Asia to detect and account for cryptic diversity in the genus.

CONCLUSION
The present study represented the first comprehensive molecular phylogenetic analysis of Dicranopteris and, therefore, provided additional insights on the Gleicheniaceae tree of life. On the basis of our molecular phylogeny from cpDNA and morphological evidence, we showed that Dicranopteris hides considerable cryptic diversity, and we were able to disentangle two new cryptic species, D. austrosinensis sp. nov. and D. baliensis sp. nov., from the diverse species complex, D. linearis. We also clarified the previously unsettled status of five taxa in Dicranopteris and, thus, proposed new combinations. We found that Dicranopteris evolved 213.04 Mya and underwent a flourishing diversification during the Miocene leading to much of its extant diversity. Our study highlighted the importance of applying morphology and molecular data to resolve relationships within taxonomically recalcitrant genera and seek out cryptic diversity. Within Dicranopteris, our results probably represented only the "tip of the iceberg" in terms of cryptic diversity, and much more intensive sampling is critical in future studies. Diagnosis: -The new species is similar to Dicranopteris taiwanensis in having a pair of accessory branches at ultimate branches, while it differs in the main forks presenting a latent bud with two lobed pseudostipules.
Description: -Plants terrestrial, 1-1.5 m tall. Rhizome creeping. Rachis stamineous, moderately brown scales, ca. 30-50 cm long between pinnae, 2 or 3 times dichotomously branched, primary rachis-branches almost equally forked, first branch 2.5-3.8 cm long and 1.0-1.8 mm wide, second branch shorter, 1.8-2.2 cm long and 0.8-1.0 mm wide, opposite branches of equal length, accessory branches at each dichotomy except for the ultimate branch, distal branches with overlapping segments. Rachis buds covered with short and usually concolorous red-brown scales. Dormant buds ovate, ca. 3-5 × 8-10 mm, covered with dense red-brown hairs, pseudostipules lanceolate. Pinnae lanceolate, generally large, ultimate pinnae longest, 25-30 cm long, 8-10 cm wide, basal pairs of segments slightly narrowed, apex acuminate, 7-8 pairs of distal segments decussate, with red-brown hairs abaxially.  Notes: -Dicranopteris inaequalis is closely related and similar to D. subspeciosa, but is recognized by having very unequal lateral branch-systems. D. inaequalis often dichotomously forks 4 or 5 times and angle of the secondary and later rachis forking more than a right angle. The smaller branches at once forked occasionally lack an accessory branch. Pinnae are lanceolate or broadly lanceolate and have 2-4 pairs of basal decrescent segments. Segments are 1.5-2 mm wide, 1.5-2 cm long, and covered with pale hairs.

Further Taxonomic Novelties
Distribution and ecology: -Dicranopteris inaequalis is mainly distributed in Malay Peninsula. It is locally common along roadsides, slopes and forest edges at ca. 1200-1400 m.
Notes: -Dicranopteris alternans is a relatively little common species. It is similar to D. subpectinata, and the most important difference between them is the indument on the lower surface of lamina. In D. alternans the lower surface of lamina bears copious red-brown hairs, whereas quite glabrous in D. subpectinata. D. alternans has rather firm lamina as well as broad-lanceolate and large pinnae with attenuate apex. Ultimate pinnae are 20-30 cm long and 4-5 mm wide. Lateral branches usually fork more than three times, and the branches at successive forks are alternately unequal. ≡ Dicranopteris linearis var. subspeciosa Holtt., Reinwardtia 4 (1957) 278.
Notes: -Dicranopteris subspeciosa is also an uncommon species. It is characterized by having alternately unequal branches with the angle of the ultimate fork being 90 • or less than 90 • . The basal two branches are of roughly similar length that may be up to twice of the same length as the ultimate branches. The species has the smallest pinnae, 13-15 cm long, 4-5 cm wide. The pinnae are long-lanceolate with caudate apex. Typically, D. subspeciosa has pale, slender and entangled hairs covering the lower surface of lamina.
Distribution and ecology: -Best known in Malaysia, where it grows along roadsides and on the edge of forest, but also known from scattered specimens from Philippines.
Specimen examined: -Malaysia. Borneo, Sabah, Kiau, Mt Kinabalu, October 31, 1915, Topping, D.L. 1516October 28, 1915, Clemens, MS 9789 (MICH, photo Notes: -Dicranopteris latiloba is one of the species with the widest segments (ca. 5-6 mm) in the genus. It is charactered by lacking a pair of accessory branches at ultimate forks, by having deeply and broadly lobed segments and quite glabrous lower surface of lamina. Usually, D. latiloba is dwarf, lateral branches forked about 1-3 times. Branches at all forks are subequal with somewhat irregularly placed segments. Segments are oblong or emarginate, firm but not very thick, reflexed when dry. Typically, the angle of ultimate fork with no accessory branches is approximately a right angle.
Distribution and ecology: -It is locally common in Malaysia and Philippines. Growing in montane forests at 1000-1500 m. Notes: -This name was coined by Copel based on a collection from Curran. Our study has convinced us that D. curranii is conspecific with D. gigantea named by Ching et al. (1959). Because D. curranii was published before D. gigantea, we therefore consider that D. gigantea should be treated as a synonym of D. curranii. D. curranii is larger as compared to any other species in the genus. Rachis dichotomously fork 2 or 3 times, with a pair of lateral pinnules at each dichotomy. It has the longest and widest pinnae. Ultimate pinnae are commonly 40-60 cm long and 13-15 cm wide, lanceolate, long caudate apex. Segments are 5-8 cm long, 3.5-4.5 mm wide, linear, reflexed when dry. Lower surface of costae and costules present sparse, brown, and deciduous hairs.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material.

AUTHOR CONTRIBUTIONS
ZYW and YHY designed the experiments. ZYW, ZQX, and JPS performed the research. HS, XLZ, WX, BA, and YHY collected the materials. ZYW and ZQX analyzed the data. ZYW and LJC drew the figures and wrote the draft. QY processed the figures. YHY, SM, and JGC revised the draft. All authors contributed to the article and approved the submitted version.