1. Introduction
The PPG classification, a comprehensive phylogeny-based classification of pteridophytes (including ferns and lycophytes), was established by the Pteridophyte Phylogeny Group in 2016 [1] , and this classification has been widely accepted and cited in recent years. Based on the PPG classification, eupolypods, composed of two clades: eupolypods I and eupolypods II, include nearly 6000 species—more than half of extant fern diversity [1] . However, there are two different hypotheses on the divergence time for eupolypods based on molecular dating: Cretaceous diversification or a Jurassic one. Most DNA-based divergence-time studies indicate that eupolypods diversification occurred in the Late Cretaceous [2] [3] [4] [5] [6] , but more recent studies demonstrated that eupolypods originated as early as in the Jurassic [7] [8] [9] . However, none of these multiple published DNA-based divergence-time studies utilized both morphological and molecular data, much less that have integrated both types of data into a total evidence (tip-dating) approach. This tip-dating approach allows for the inclusion of morphological data in divergence-time estimates based on both extant and extinct taxa.
So far there are three different approaches to date the diversification time for an organism group. At the very beginning when calibrating a tree, the rate-dating (clock rate) method was applied; subsequently, researchers used to date the tree based on node calibrations (node-dating), and finally, researchers used total-evidence dating (TED or tip-dating), in which the tree is dated based on simultaneous analysis of fossil and recent taxa [10] [11] [12] . Compared to the first two dating methods, the tip-dating approach of phylogenetic divergence-time estimations has been revolutionized by two recent developments: 1) the approach of tip-dating allows for the incorporation of fossils as tips in the analysis, with their phylogenetic and temporal relationships to the extant taxa inferred from the data, and 2) the fossilized birth-death (FBD) class of tree models that capture the processes that produce the tree (speciation, extinction, and fossilization) and thus provide a coherent and biologically interpretable tree prior [13] . In this study, we apply an integrative tip-dating approach that combines molecular and morphological data to re-estimate divergence times for eupolypods—the most diverse of all major lineages of ferns in the light of the stratigraphic records. In this study, we will apply the new molecular dating approach to attempt to answer the question of what geological time eupolypods start to diversify from, thereby increasing our understanding of its evolutionary history.
2. Analysis Methods
2.1. Sampling Taxa Set and Assembling Dataset
So far there are some studies on divergence times reported to date the diversification time for eupolypods, after reviewing the main and influential studies of fern phylochrological analyses (Figure 1, Table 1), we found only the analyses of Regalado et al. [8] focused on relationships among eupolypods lineages. To evaluate the possible relationships between fossil Holttumopteris burmensis Regalado et al. [8] and extant eupolypods lineages, Regalado et al. [8] mapped character states of the fossil H. burmensis on a reconstructed comprehensive, time-calibrated phylogeny of eupolypod ferns. Therefore, we used those two datasets of Regalado et al. [8] , both morphological character matrix and DNA sequence matrix can be downloaded from the online resource of Regalado et al.
![]()
Figure 1. The divergence time estimates for eupolypod ferns (clades in red lines) are simplified from previous main studies for fern phylochrological analyses (Table 1). Mean ages in brackets after references are shown on the right side. Black arrows at nodes indicate the clades of eupolypod ferns where the mean ages are shown. Stars with lines in red indicated that the diversifications of eupolypods occurred in the Cretaceous, and green stars and lines indicated that the diversifications of eupolypods dated back in the Jurassic. Two main apomorphic characters of eupolypod ferns (sporangia with vertical annulus interrupted by the sporangium stalk and monolete spores) are shown on the right side too.
![]()
Table 1. Summary of fern phylochrological analyses in previous studies and this study.
[8] . We added nine fossil taxa and six extant taxa to that combined dataset. Fossils included in this study and their information are shown in Table 2. The six additional extant species are from Desmophlebiaceae (one taxon), grammitid ferns (two taxa), and Drynaria (three taxa) that were not included in the dataset of Regalado et al. [8] . Their DNA sequences were downloaded from GenBank, and their GenBank accession numbers are followed by the species name and are shown in the chronogram of eupolypod ferns (Figure 2). For the data matrix of morphological characters, four more characters are added to the dataset of Regalado et al. [8] , which includes annulus cell number of a sporangium, vascular bundle number of a petiole, spore types based on [14] , and frond dimorphism. Subsequently, we combine DNA sequence data with morphological character data from 9 fossil and 209 extant taxa (2003 taxa of eupolypods and four taxa of basic polypods as outgroup members) to simultaneously infer the phylogeny and re-estimate divergence times for eupolypod ferns by a total-evidence (tip-dating) approach. We produced a matrix of 15 morphological characters coded for a
![]()
Table 2. Fossils included in this study and their information.
sample of 218 fern species, the taxa were chosen to represent the phylogenetic diversity of the family level following Regalado et al. [8] . We also used sequences of three plastid genes (rbcL, atpA, and atpB; 3841 aligned positions).
2.2. Set Calibration Priors for Bayesian Tip-Dating Analyses
Bayesian tip-dating analysis using the fossilized birth-death process (FBD) was previously conducted only for one fern lineage, the Marattiales [13] . They fitted a variety of complex, hierarchical Bayesian models that varied in assumptions regarding how morphological characters evolve, how evolutionary rates vary across the tree, and whether rate shifts are coincident with geologic intervals [13] . May et al. [13] used stepping-stone analysis [15] to estimate marginal likelihoods for each model to determine the best-fit model. We applied the same best-fit model configuration as determined by [13] to our updated character data matrix. Our tip dating analyses were performed in Mrbayes 3.2.7a [11] [16] following the manuals downloaded from http://mrbayes.net. The FBD model was used as the tree prior, the ages of fossil terminals were provided as uniform or fixed priors with bounds equal to the limits of the estimated ages of their deposits (Table 2). We used a “diversity” setting in the sampling strategy since we strived to include as many eupolypods terminals as possible, and set the sample probability prior to 0.0178; this was done because we included 214 terminals, while the diversity of the eupolypods lineage is currently of about 6000 species and we expected it contains about that same number of undescribed species, so, 214 (eupolypods taxa this study) divided by (6000 + 6000, the total diversity of the eupolypods) is equal to 0.0178. Our analyses were run for 20 million generations, sampling every 5000 generations. Bayesian posterior probabilities (PP) were calculated for the majority rule (>50%) consensus tree of all sampled trees after discarding the first 25% as burn-in. Visualized trees and all the nodes were checked in FigTree v.1.4 [17] .
3. Results
3.1. Tip-dating Analyses
Tip-dating analyses of the total-evidence dataset (combined 3841 molecular characters and 15 morphological characters) resulted in a total of 214 eupolypod taxa falling into two separate clades: eupolypod I with 126 taxa and eupolypod II with 78 taxa (Figure 2). The age estimation for the diversification of eupolypods is in Jurassic, mean 152.15 Ma, with a 95% highest posterior density (HPD) interval of 128.76 - 178.80 Ma, the diversification of eupolypod II also occurred in
![]()
![]()
Figure 2. The chronogram of eupolypod ferns using tip-dating approach with the Fossilized Birth-Death model in a Bayesian framework. Nine fossil taxa are in red, six species in green with a Genbank accession number are the extra taxa added to the dataset [8] . Numbers above the branches indicate posterior probability (pp%, only shown over 50), numbers below the branches indicate the mean age (Ma, only shown for the main clades), and the nodes where the main lineages of eupolypods diversified are marked with a star. SG, stem group; CG, crown group. (A) the Eupolypods II part, (B) the Eupolypods I part.
Jurassic, mean 145.29 Ma, with a 95% HPD interval of 121.05 - 165.88 Ma, while eupolypod I diversified later, which is in Cretaceous, mean 102.44 Ma, with a 95% HPD interval of 71.96 - 136.22 Ma (Figure 2).
3.2. Phylogenetic Positions of Fossils
Our tip-dating tree shows that the phylogenetic placements of fossils are mostly in accordance with their previous taxonomic allocations, among the nine fossil records, six fossils are supposed to belong to eupolypod II based on previously published morphologic descriptions and their taxonomic status, and as be predicted based on morphological studies, they all occupied the stem positions of eupolypod II lineage in our tip-dating tree (Figure 2). Three fossils are supposed to belong to eupolypod I are all in the lineage of eupolypod I. However, only three fossils of eupolypod I have more concrete and accurate phylogenetic placements, for an example, the fossil Davallia walker [21] is clustered with its extant Davallia relatives (Figure 2).
We suppose that providing a more concrete and accurate phylogenetic placement for those fossils that are dated back as old as the origin of the whole lineage is challenging, because tip-dating analyses allow the fossil age to inform about their phylogenetic hypothesis. More fossils as well as larger DNA sequence datasets for extant taxa may compensate for the limited number of morphological characters identifiable on fossils, serving as a backbone for reliably estimating the phylogeny of both extinct and extant taxa.
4. Discussions
4.1. Diversification of Eupolypods
So far, there have been eight main and influential studies of fern phylochrological (divergence time estimation) analyses related to the diversification of eupolypods, five of the eight previous studies supported a Cretaceous diversification of eupolypods, the more recent three studies supported a Jurassic diversification of eupolypods (Table 1), our tip-dating analyses supported the second hypothesis. This result is nevertheless important because it provided another estimation based on the recently developed tip-dating analyses that eupolypods originated before the Cretaceous, as early as the Jurassic. This result is consistent with the few divergence time estimates based on the different molecular dating methods (node-dating [7] [9] and rate-dating [8] ). On the other hand, the fossil record has always been considered as crucial information as an independent source of evidence for diversification of eupolypods. Recent studies on the Myanmar amber fossils (late Albian-earliest Cenomanian, Cretaceous, c. 100. 5 Ma) have recorded remarkable fossils—both lineages of eupolypods has at least one representative fossil: eupolypod II with the fossil of Thelypteridaceae [8] and eupolypod I with fossil of Dryopteridiaceae [28] [29] . Both fossils mentioned above and newly available fossils (scales and spores assigned to eupolypods [30] ) further substantiate that Polypodiales were more diverse and abundant in the mid-Cretaceous forest ecosystems of Myanmar clearly showing that eupolypods originated before the mid-Cretaceous, probably as early as the Early Jurassic [30] , this is consistent with the few divergence time estimates based on other molecular dating studies [7] [8] [9] and our tip-dating analyses.
4.2. Bayesian Tip-Dating
Extinct organisms, i.e., the fossils, could provide crucial information about the origin and time of origination of extant groups, however, in previous studies (Table 1), stratigraphic age ranges of fossils were used to calibrate molecular clock analyses only (the traditional node-dating method), morphological characters, however, were not directly incorporated. The very early rate-dating method, using the standard substitution rate of nucleic acids or amino acids, did not use the information provided by the fossils at all. The importance of morphological phylogenetics for rigorously dating the Tree of Life is now widely recognized and has been revitalized by methodological developments such as the application of tip-dating Bayesian approaches [31] . The introduction of tip calibrations allows fossil species to be included alongside their living relatives, which overcomes several drawbacks in the traditional stepwise approach to some extent, however, it also requires more knowledge about statistics and more priors need to be done when running the program. How to set the priors for the divergence times, the priors for the evolutionary rates, and the models for the character changes that are all key components in this method, probably continue to be a significant challenge for its users [16] .
5. Conclusion
For the first time, this study combined morphological data from extinct and extant taxa with DNA sequence data to estimate diversification ages for eupolypods, the most diverse of all major lineages of ferns. Not only did we obtain the age estimates, but also the phylogenetic positions of fossils in one tip-dating analysis. The results further strengthen previous Jurassic diversification hypotheses for eupolypods. This is undoubtedly a successful experimentation. Hence, this study unequivocally validates fossil tip-dating as a promising tool within a phylogenetic framework. To bring more details regarding the fern evolutionary history to light, more studies using tip-dating methods are currently underway.
Acknowledgements
The authors wish to thank the editors and anonymous reviewers for constructive suggestions and comments that have improved the paper. This research was supported by the Basic Frontier Scientific Research Program of the Chinese Academy of Sciences (No. ZDBS-LY-DQC021-02). The authors deeply appreciate the support.