1. Introduction

In terms of species diversity and endemism, the Indo-Burma Biodiversity Hotspot—which comprises all non-marine parts of Cambodia, Lao, Myanmar, Thailand, and Vietnam, plus parts of southern China—is one of the most biologically important regions of the planet (Tordoff et al., 2012). It is one of the 25 biodiversity hotspots recognized by Myers et al. (2000). Geographically, this hotspot is located among India, the Himalaya, southern China, and the Sundaic Region, and thus its biota is a mixture of the floras and faunas of these regions, and has a significant endemic component, particularly in the case of plants (Tordoff et al., 2012). Van Dijk et al. (2004) estimated the total plant diversity in the hotspot to be about 13, 500 vascular plant species, of which about 7000 (52%) are endemic. This is probably a very conservative estimate since the diversity of many plant groups has not been well explored. One of these plant groups that deserve more attention is the Old-World fern genus Leptochilus Kaulf. (Polypodiaceae; Figs. 1 and 2).

The java fern genus Leptochilus, as defined by Zhang et al. (2019), holds particular significance in exploring the plant diversity in the Indo-Burma Biodiversity Hotspot. When established in 1824, Leptochilus initially included only one species. Mainly based on molecular evidence, four genera, Colysis C. Presl, Dendroglossa C. Presl, Kontumia S.K. Wu & L.K. Phan, and Paraleptochilus Copel., were treated as synonyms of Leptochilus (Dong et al., 2008; Kreier et al., 2008; Kim et al., 2013; Zhang et al., 2019). Despite this, Leptochilus was commonly recognized as a small genus with 25 "indistinct" species (Zhang and Nooteboom, 2013). Recently, new species and combinations have been continuously added to Leptochilus, partially increasing the species number of the genus (Zhang et al., 2015, 2018, 2023, 2024; Zhao et al., 2017; Liang et al., 2020; Fujiwara et al., 2023; Wei et al., 2023; Yu et al., 2024). The currently defined Leptochilus is distributed from East Asia, South Asia to Southeast Asia (Fig. 1). The genus is distinguished within Polypodiaceae by diverse characteristics, including various shapes of sori (orbicular, elongate, linear, acrostichoid), frond dimorphism (monomorphic, dimorphic), and lamina dissection (simple to pinnatifid, 2-pinnate to 4-pinnatifid in L. heterophyllus; Fig. 2).

Based on molecular data of the chloroplast markers, recent phylogenetic analyses have uncovered distinct genetic differences that exist not only among species with different morphologies but also within those species morphologically similar to one another (Chen et al., 2020, 2023; Yu et al., 2024). For example, species with pinnatifid laminae and linear sori, often referred to as the Leptochilus ellipticus [all species authorities are listed in Appendix A] complex, were resolved in three of the six recognized clades in Leptochilus (Zhang et al., 2019). This underscores the prevalent occurrence of convergent evolution in diagnostic morphological traits within the genus. Zhang et al. (2019) also pointed out the presence of at least 10 cryptic species to be recognized in their reconstructed phylogeny. Due to limited information about certain core species, particularly their phylogenetic relationships, the identities of almost half of the species in the genus remain uncertain. For example, within the L. ellipticus complex, many specimens with large habit and lobes were identified as Leptochilus pothifolius. However, as Fraser-Jenkins (2008) indicated, the typical L. pothifolius possesses 2–3 large lobes, suggesting that a considerable number of specimens labeled as this species might be due to the misinterpretation of the species' type specimens. Regarding specimens displaying large lobes similar to those of L. pothifolius but possessing 3 pairs of lobes, confidently recognizing additional species becomes difficult until a more comprehensive comprehension of L. pothifolius, particularly its phylogenetic relationships, is acquired.

In this study, we used an updated phylogeny with additional samples collected in the past decade to further elucidate the relationships within Leptochilus and, in conjunction with morphological and biogeographical studies, to further explore the diversity and evolution of the genus.

2. Materials and methods 2.1. Taxonomic sampling and DNA sequencing

To determine the phylogenetic relationships and to explore the species diversity in Leptochilus, we reconstructed a large plastid phylogeny with the most comprehensive sampling of the genus to date by doubling previous largest taxon sampling so far (Zhang et al., 2019) by including as many samples as possible. In addition to 121 newly added samples of Leptochilus, 105 samples of the genus sequenced in previous studies were included. Those 226 samples, representing approximately 70 species (including 21 undetermined species), were collected from 11 Asian countries. Microsorum punctatus was used to root the tree. Species identification was conducted by combining information from field observations, morphology of the type specimens, geographic distribution, and phylogenetic results. When identification based on morphology contradicted the phylogenetic results, a 'cf.' was added before the epithet to indicate the uncertainty of the identification. Updated identifications compared to those in previous studies are listed in Appendix A.

Six plastid markers (atpB, rbcL, rps4, the rps4-trnS intergenic spacer, the trnL intron, and the trnL-F intergenic spacer) were selected for amplification and sequencing because those markers are easily to be amplified and their resolving power for phylogenetic studies of microsoroid ferns (Kreier et al., 2008; Zhang et al., 2019; Chen et al., 2020). For nuclear phylogeny, pgiC gene was chosen. The nuclear gene pgiC was amplified with 14F and 16R primers of Ishikawa et al. (2002). Most nuclear pgiC data were obtained via direct sequencing, although cloning was necessary in some cases when direct sequencing failed. Cloning was carried out using the pEASY-T3 Cloning Kit (TransGen Biotech, Beijing, China), following the manufacturer's protocols; mostly five colonies per accession were sequenced.

Total genomic DNA was extracted from silica-dried material using the TIANGEN plant genomic DNA extraction kit (TIANGEN Biotech., Beijing, China) following the manufacturers' protocols. Primers and PCR conditions for plastid genes followed Zhang et al. (2019).

2.2. Sequence alignment and phylogenetic analysis

The sequences downloaded from GenBank and the newly generated data were aligned using MAFFT (Katoh and Standley, 2013) with subsequent manual adjustments with BioEdit (Hall, 1999). Maximum likelihood tree searches and ML bootstrapping were conducted using RAxML-HPC2 on TG v.7.2.8 on CIPRES platform (Miller et al., 2010; Stamatakis et al., 2008), with 1000 rapid bootstrap (BS) analyses followed by a search for the best-scoring tree in a single run (Stamatakis et al., 2008). To further analyze the combined dataset, parsimony jackknife (JK) analyses were conducted using PAUP∗ with indels treated as missing data, the removal probability set to approximately 37%, and "jac" resampling emulated. One thousand replicates were performed with 10 TBR searches per replicate and a maximum of 100 trees held per TBR search. Bayesian inference (BI) analyses were executed using MrBayes v.3.1.2 (Ronquist and Huelsenbeck, 2003) on the CIPRES platform (Miller et al., 2010). Two independent runs were performed, each employing four chains (one cold and three heated) with a temperature parameter set to 0.2, a transition/transversion rate ratio set to beta, and priors set to their default values. Each run initiated with a randomly generated tree, and trees were sampled at intervals of 1000 generations over a total of 10, 000, 000 generations. To ensure convergence and stationarity, assessments were conducted using Tracer v.1.4 (Rambaut and Drummond, 2007). The first 25% of the generated trees were discarded as burn-in, ensuring the attainment of stationarity in log-likelihood. The remaining trees were then utilized to derive a 50% majority-rule consensus topology and calculate posterior probabilities (PP). For the nuclear phylogeny, only ML analysis was performed using the methods described above.

2.3. Divergence time estimation

To estimate the divergence of the Leptochilus, we limited the sampling to contain one or two accessions for each species. The reduced alignment contained 87 accessions. We used BEAST v.1.10.4 (Drummond and Rambaut, 2007) to run the dating analysis, employing the uncorrelated lognormal clock model with the GTR + G nucleotide substitution model as suggested by Partitionfinder v.2.1.1. Birth-death speciation was specified for the dataset, with a uniform prior set to 0 to 10 for growth rate and 0 to 1 for relative death rate; initial values were set at 1 for growth and 0.5 for relative death. Due to the absence of Leptochilus fossils, we employed three secondary calibrations from Du et al. (2021) and Testo and Sundue (2016). The root age was constrained to 37.77 Ma with a uniform prior of 34.6–43.53 Ma. The stem and crown ages of Leptochilus were assigned as 25.96 and 19.62 Ma, respectively, with lognormal priors for offset and standard deviation set to 1.0. We conducted two runs of four Markov chain Monte Carlo chains with 100 million generations, sampling every 5000 generations. The effective sample size (ESS) was examined using Tracer 1.7 (Rambaut and Drummond, 2007) to ensure asymptotic behavior of the likelihood values, with ESS > 200. LogCombiner v.1.10.4 and TreeAnnotator v.1.10.4 were used to obtain a consensus tree, discarding the first 2000 trees as burn-in and combining the remaining 98, 001 trees to generate a maximum credibility tree.

2.4. Biogeographical analyses

The distribution of each species was determined based on a monograph by Nooteboom (1997), a recent taxonomic work (Zhang et al., 2024), various floras (e.g., Fraser-Jenkins, 2008; Zhang and Nooteboom, 2013), and the online data bases, including Chinese Virtual Herbarium (CVH, http://www.cvh.ac.cn), Global Biodiversity Information Facility (GBIF, https://www.gbif.org), and Pteridophyte Collections Consortium (PCC, https://www.pteridoportal.org/portal/). Three geographical regions, Malay Archipelago, Indo-Burma, and East Asia (Fig. 6), were divided based on the distribution pattern of Leptochilus. The dated phylogeny generated in BEAST was used as input for the subsequent biogeographical analyses. The biogeographical reconstructions were conducted using BioGeoBEARS (BioGeography with Bayesian (and likelihood) Evolutionary Analysis in R Scripts; Matzke, 2014), as implemented in RASP 4 (Yu et al., 2020). The DEC models (Ree and Smith, 2008), DIVALIKE, and BAYAREALIKE, each with a factor 'j', were compared using likelihood approach to generate biogeographical events and ancestral areas (Matzke, 2013).

2.5. Morphological analysis

Field observations were conducted during our fieldwork in China, Laos, Nepal, Thailand, and Vietnam. Characteristics such as scale colors, and cross sections of laminae and petioles, were observed using a microscope. Specimens of Leptochilus from the herbaria A, CDBI, K, KUN, L, MO, P, PE, PYU, SING, TAIF, and VNMN were examined. In total, 3630 specimens were examined (Fig. 1). Three major morphological characters, lamina shape (simple and entire, trifid or palmate, pinnatifid, 2-pinnate to 4-pinnatifid), frond dimorphism (dimorphic, monomorphic), and soral shape (acrostichoid, linear, irregularly rounded or elongate) were analyzed. Mesquite v.2.75 (Maddison and Maddison, 2008) is used to optimize characters on the most likely tree inferred from ML analysis. For those species with multiple samples, only one or two samples were retained in the tree for analysis. Characters were unordered and equally weighted. The lamina of L. pteropus was coded as simple and entire, though it can vary from irregularly lobed to trilobed. The soral shape of L. hemionitideus and L. macrophyllus was coded as linear, though the sori can be interrupted into shortly linear and elongate. Despite some individuals of L. multilobus have elongate sori, the soral shape of this species was coded as linear, given the overall soral appearance.

3. Results 3.1. Phylogenetic relationships

A total of 190 plastid sequences and 47 nuclear ones were newly generated for this study (Appendix A). The alignment of six plastid markers was 5965 characters long, of which 452 sites were identical, 1086 characters were parsimony-informative, and 408 variable characters were parsimony-uninformative. The alignment of the nuclear pgiC data was 779 characters long, of which 257 sites were identical, 138 characters were parsimony-informative, 110 variable characters are parsimony-uninformative. An examination of the trees generated in MPJK analyses for individual plastid markers, as well as the combined plastid dataset, revealed no well-supported conflicts (MPJK 70%; Mason-Gamer and Kellogg, 1996; Zhang and Simmons, 2006). However, the resolution of the Leptochilus ellipticus clade based on the nuclear data was in conflict with that based on plastid data. Thus, nuclear data and plastid data were analyzed separately.

Our plastid phylogeny of Leptochilus was generally consistent with previous studies (Zhang et al., 2019; Chen et al., 2020). The monophyly of the genus was strongly supported. Within Leptochilus, the six major clades identified by Zhang et al. (2019) were recovered and they were all strongly supported as monophyletic except the L. ellipticus clade (MLBS: 61%, MPJK: < 50%, BIPP: 0.95). In addition to these six clades, we found three new major clades in the plastid phylogeny with their relationships not well supported: clade I, clade II, and clade III. The former two containing three deeply diverged species were weakly supported as monophyletic and they together were weakly supported as sister to the L. ellipticus clade. Clade III contained three deeply diverged species and was weakly supported as sister to the Colysis clade. The nine major clades were sequentially from the bottom to the top of the tree the L. pteropus clade, the L. macrophyllus clade, the Vietnam clade, the L. pentaphyllus clade, clade III, the Colysis clade, clade II, clade I, and the L. ellipticus clade. The relationships among the L. pentaphyllus clade, clade III, the Colysis clade, clade I + II, and the L. ellipticus clade were weakly (< 50%) supported. The relationships between the L. pteropus clade and the rest were only weakly to moderately (MLBS: 76%, MPJK: < 50%, BIPP: 0.71) supported. The relationships among other major clades were strongly supported.

In addition to the three major clades in the plastid phylogeny, we also found four new subclades in the Colysis clade and two new subclades in the Leptochilus ellipticus clade. The six subclades each contained one species currently unknown to science (Fig. 3).

The overall resolution in the nuclear phylogeny was poor but not in conflict with that in the plastid phylogeny except the Leptochilus ellipticus clade which was strongly supported as paraphyletic in relation to the Vietnam clade. Other conflicts were not well supported (Fig. 4).

The occurrences of 3630 specimens of Leptochilus examined are shown in Fig. 2 and the type localities of the 51 currently recognized species including 11 just described (Zhang et al., 2024) are indicated in green dots.

3.2. Divergence times analyses

The reduced and dated tree showed the same topology as the Maximum Likelihood phylogeny based on the complete sampling (Fig. 3). BEAST analyses estimated a putative divergence of Leptochilus from other microsoroid ferns around 27.3 Ma (95% HPD: 25.7–29.2 Ma), and diversified during the early Miocene, 22.8 Ma (95% HPD: 21.2–24.5 Ma). As one of the earliest diverged lineages, the L. pteropus clade was diversified around 12.8 Ma (95% HPD: 7.1–18.4 Ma). The L. macrophyllus clade split from its closest relative ca. 22.1 Ma (95% HPD: 20.2–23.9 Ma), with the first diversification at 11.1 (95% HPD: 6.6–16.1 Ma). The divergences of the other seven clades were estimated to be at 20.3 Ma (95% HPD: 18.0–22.6 Ma), following the radiation of the Vietnam clade at 6.2 Ma (95% HPD: 2.4–12.0 Ma), the Colysis clade at 13.7 Ma (95% HPD: 10.9–16.5 Ma), and the L. ellipticus clade at 13.5 Ma (95% HPD: 11.0–16.0 Ma). Of the three new clades, clade Ⅰ diverged from clade Ⅱ at 9.4 Ma (95% HPD: 5.2–13.7 Ma), whereas clade Ⅲ diverged from the L. pentaphyllus clade at 14.2 Ma (95% HPD: 10.9–17.4 Ma; Fig. 5).

3.3. Ancestral area reconstruction

The analysis of ancestral area reconstruction identified BAYAREALIKE as the best-fitting model with a significant higher AICc weigh value (AICc_wt = 0.68) compared to other models (Table 1). According to this model, the ancestor of Leptochilus was most likely distributed in the Indo-Burma and Malay Archipelago areas. Except for the L. pteropus clade and the L. macrophyllus clade, whose ancestors probably originated from the Indo-Burma and Malay Archipelago areas, the ancestor of all other seven clades originated from the Indo-Burma region. Sixteen dispersal events were detected, including one from EAST Asia to the Malay Archipelago, one from the Malay Archipelago to East Asia, and 14 from the Indo-Burma region to East Asia (Fig. 6).

3.4. Evolution of morphological characters

Character states of three important characters, lamina shape, frond dimorphism, and soral shape, are mapped on the plastid tree (Fig. 3). The reconstructions of three morphological characters are shown in Figs. 7-9. It took 16 steps for the parsimony reconstruction of lamina shape, 7 steps for frond dimorphism, and 10 steps for soral shape. The ancestral state of frond dimorphism was monomorphic, the ancestral state of lamina shape could either be simple and entire or pinnatifid, and the ancestral state of soral shape was rounded or elongate.

4. Discussion 4.1. New findings in the plastid phylogeny and the nuclear phylogeny

According to the molecular data from six plastid makers obtained from 105 accessions of Leptochilus, Zhang et al. (2019) conducted the first comprehensive study on the phylogeny of the genus, proposing six clades and 13 subclades within Leptochilus. In this study, using the same markers, we reconstructed the phylogeny with 226 samples of Leptochilus. With no well-supported conflicts found when comparing with results of Zhang et al. (2019), we here identified three new major clades and six subclades (Fig. 3).

4.1.1. Three new clades

Three new clades were resolved as a polytomy in relation to the Leptochilus ellipticus clade, the Colysis clade, and the L. pentaphyllus clade, in spite of the high genetic variations among those clades. Given that all the species of those clades have pinnatifid laminae similar to most species in the L. ellipticus clade, we infer that the three new clades, along with L. pentaphyllus clade, could indeed belong to the L. ellipticus clade in a well-resolved phylogeny. Among the three new clades, L. gracilis of clade Ⅰ and L. bolikhamsaiensis of clade Ⅲ are distinguished by their smaller habits compared to other species with pinnatifid laminae, and were each resolved as sister to a species with medium habits. Clade Ⅱ, represented by L. pothifolius, a Himalayan endemic species with pinnatifid laminae, was genetically divergent to the morphologically similar species distributed in Southeast Asia or East Asia.

4.1.2. Six new subclades

Among the six new subclades, subclades A and B within the Leptochilus ellipticus clade were each represented by a recently described species, L. ornithopus and L. scandens, respectively (Fig. 2). Those two species are distinguished from other species in Leptochilus by having hemiepiphytic habitat (Fujiwara et al., 2023; Wei et al., 2023), and were found to be genetically distant from their sister taxa. Subclade C, represented by L. kachinensis, a species with strongly dimorphic leaves and acrostichoid sori, is resolved as sister to the Colysis subclade formed by the L. hemionitideus complex, L. kachinensis, and L. hemionitideus that share similarities in their lamina texture and venation pattern. The remaining three subclades, D, E, and F, have elongate or linear sori, are distinguished within the Colysis clade, considering that the majority of species in the Colysis clade have acrostichoid sori. Subclade F, represented by three samples of L. multilobus from southern Xizang, is resolved as one of the earliest diverging lineages in the clade, and is currently the only member with pinnatifid laminae in the Colysis clade. Based on the tree topology, we can infer that the pinnatifid lamina of subclade F is derived from simple lamina.

4.1.3. Nuclear phylogeny

For pteridophyte phylogeny, available good nuclear markers with sufficient evolutionary rates for inter-specific relationships are quite scarce and thus there have been very few infrageneric nuclear phylogenies on pteridophytes published (e.g., Nitta et al., 2011; Sessa et al., 2012; Chang et al., 2013; Hori et al., 2014, 2016, 2018; Rothfels et al., 2014; Weststrand and Korall, 2016; Zhou et al., 2016; Fujiwara et al., 2018; Hori, 2018; Zhang and Zhang, 2018; Hori and Murakami, 2019; Liang et al., 2019; Chao et al., 2022; Zhou and Zhang, 2023). Here, we reconstructed the first nuclear phylogeny of Leptochilus. Although the overall resolution is poor and supported values in the tree are low, the nuclear phylogeny can tell us: (1) Leptochilus is strongly supported as monophyletic; (2) the L. pteropus clade and the Vietnam clade are each strongly supported as monophyletic, consistent with the plastid phylogeny; (3) the monophyly of the Colysis clade and Clade Ⅲ each are uncertain; and (4) the overall nuclear phylogeny in Leptochilus is not in conflict with the plastid phylogeny except the resolution of the L. ellipticus clade that is not monophyletic because the Vietnam clade is strongly supported as sister to portion of the L. ellipticus clade (Fig. 4), which points to possible hybridization between species in the two clades. More studies are needed.

4.2. Circumscriptions of three core species and discovery of hidden diversity

As noted by Zhang et al. (2019), the species number of Leptochilus has been underestimated partially due to the cryptic speciation. Several widely distributed species, such as Leptochilus decurrens, L. ellipticus, L. macrophyllus, L. pedunculatus, and L. pothifolius, may be composed of multiple species. Therefore, it is imperative to morphologically compare samples from different geographical regions with those collected from the type localities. In this study, we incorporated L. pedunculatus and L. pothifolius in our new phylogeny. According to the tree topology and genetic variations, approximately 55 species can be recognized, including 4 species in the L. macrophyllus clade, 2 species in the L. pteropus clade, 2 species in the Vietnam clade, 2 species in the L. pentaphyllus clade, 12 species in the Colysis clade, 32 species in the L. ellipticus clade, and 1 species without assignment of any existing clades. Based on our field observations, morphological study, geographical distribution, ecology, and molecular phylogeny, here we re-circumscribed L. pothifolius, L. pedunculatus, and L. ovatus. This re-circumscription, in turn, has facilitated the discovery of 11 new species that have been described elsewhere (Zhang et al., 2024).

4.2.1. The recognition of Leptochilus pothifolius

As pointed out by Fraser-Jenkins (2008), the species name Leptochilus pothifolius has been misapplied by many Chinese and Indian authors who followed Ching's (1933) misunderstanding of the species. In Ching's concept, L. pothifolius was characterized by large frond and more than four pairs of lateral lobes. Fraser-Jenkins (2008) regarded a specimen in the herbarium BM, barcode BM001038462, collected by F. Buchanan–Hamilton s.n. from Nepal, as the holotype of L. pothifolius, which was designated as the lectotype later by Mazumdar (2015). According to the type specimen, L. pothifolius is distinguished by having two pairs of lateral lobes and broadly winged rachis. Our materials, Liang Zhang et al. 2105 & 4794 (KUN), collected from Medog in the Himalayan regions, have 2–4 pairs of lateral lobes and broad rachis wings and may represent the true L. pothifolius. Our phylogenetic analysis failed to assign L. pothifolius to any of the six existing clades as defined by Zhang et al. (2019). Instead, the species, along with L. bolikhamsaiensis, the L. pentaphyllus clade, the Colysis clade, and the L. ellipticus clade, formed a well-supported clade.

4.2.2. The recognition of Leptochilus pedunculatus

Leptochilus pedunculatus, with its type specimens collected from eastern Bangladesh, is distinguishable in the genus by its subdimorphic and simple fronds, ovate-lanceolate lamina, and linear sori. The species was often thought to be widely distributed, ranging from N.E. India, Nepal, South China, to S.E. Asia (Nooteboom, 1997; Fraser-Jenkins, 2008; Zhang and Nooteboom, 2013). Zhang et al. (2019) considered four samples (L.B. Zhang 6334, 6378, 6692, 6931) collected from Vietnam to be morphologically similar to L. pedunculatus and placed them within the Leptochilus henryi subclade in the L. ellipticus clade. While Chen et al. (2020) identified one sample (Cheng-Wei Chen Wade1334) from Vietnam as L. pedunculatus and resolved it within the Colysis clade. In this study, three samples of L. pedunculatus were collected from Medog, Tibet, displaying morphological and geographical proximity to the type specimens. Our reconstructed phylogeny suggests that L. pedunculatus belongs to the Colysis clade but genetically distinct from the morphologically similar specimens collected from South China and S.E. Asia. This implies that L. pedunculatus may have a limited distribution at low elevations in the Himalayan region.

4.2.3. The recognition of Leptochilus ovatus

Copeland (1914) originally described Leptochilus ovatus based on fern collections by Brooks from Sumatra and subsequently placed it in Campium C. Presl and later in Paraleptochilus (Copeland, 1928, 1947). Because of its subdimorphic fronds, ovate-lanceolate lamina, and linear sori, this species has often been treated as a synonym of L. pedunculatus (Nooteboom, 1997). Our likelihood analysis, which sampled eight accessions of L. ovatus from southern Vietnam, southern Laos, and southern Thailand, resolved them as monophyletic and sister to a clade formed by L. decurrens, L. evrardii, L. hemionitideus, L. kachinensis, L. pedunculatus, and two undetermined species. Field observations indicate that L. ovatus frequently grows on a small trunk, exhibiting a hemiepiphytic habit, aligning with the description in the protologue as "scandent near base of small trees". The distribution of L. ovatus may be confined to tropical regions in S.E. Asia.

4.3. Origin and biogeography of Leptochilus

Our divergence time analyses indicated that Leptochilus originated in the Oligocene and diversified from early Miocene. Zhang et al. (2019) suggested that Leptochilus might have evolved at lower latitudes and progressively dispersed to and colonized higher latitudes. Our ancestral area reconstruction confirmed this hypothesis by detecting 15 dispersal events from lower to higher latitudes. Only one event from higher latitudes to lower latitudes was found: the current distribution of L. decurrens in the Malay Archipelago could be partially explained by ancestral dispersals from the Indo-Burma region. The 15 dispersal events from the Indo-Burma region to East Asia may have resulted in distribution/evolution of the following 15 species: the colonization of L. hemionitideus in East Asia; L. ellipticus in mainland China, Taiwan Island (China), and Japan; L. elegans in Taiwan Island (China) and Japan; L. sp. 1 in Guizhou, China; L. sp. 2 in Guangxi, China; L. digitatus in East Asia; L. leveillei in southern China; one species in the L. flexilobus complex (DJY04059) in Sichuan, China; L. flexilobus in southern China; L. shintenensis and L. wrightii in Taiwan Island (China); L. henryi in China; and three other dubious species in East Asia pending more studies. The Java fern L. pteropus is one of the most widely distributed species in Asia, and its colonization in South China, Taiwan Island (China), and South Japan may be attributed in its ancestral expansion from the Indo-Burma and Malay regions towards higher latitudes (Fig. 6).

4.4. Morphological evolution in Leptochilus

Lamina shape, frond dimorphism, and soral shape are three major morphological characters used in the classification of Leptochilus (Zhang et al., 2024). By mapping the states of the three characters onto our reconstructed phylogeny (Fig. 3) and tracing their evolutionary states (Figs. 7-9), we can infer the evolution of these characters.

4.4.1. Dissected laminae from simple and entire laminae

The dissected laminae in Leptochilus might have evolved from simple and entire laminae, because the two earliest-diverging clades, the L. pteropus clade and the L. macrophyllus clade, consist of species with simple and entire laminae. However, when considering the outgroups, the character evolutionary analyses suggested that the ancestral state of lamina shape can either be simple and entire or simple and pinnatifid (Fig. 7). Within the Colysis clade, the pinnatifid laminae of L. multilobus may have evolved from simple and entire laminae, as all other species in the clade have simple and entire laminae. L. heterophyllus, the only species with strongly dimorphic fronds and bipinnate to 4-pinnatifid laminae in Polypodiaceae, was weakly supported as a member of the L. ellipticus clade. The highly dissected laminae of L. heterophyllus may have evolved directly from simple and entire laminae or from simple and pinnatifid laminae. Analogously, Selliguea dareiformis (Hook.) X.C. Zhang & L.J. He (Polypodiaceae), which has monomorphic fronds and 3- or 4-pinnate laminae, is phylogenetically nested within the lineage formed by species with simple or pinnate laminae in that genus (Schneider et al., 2002).

4.4.2. Strongly dimorphic fronds from monomorphic fronds

Sterile-fertile leaf dimorphism is a special case of morphological diversity of fern leaves (Vasco et al., 2013), potentially improving the spore dispersal ability and nutrient distribution. Given that the three earliest-diverging clades, the L. pteropus clade, the L. macrophyllus clade, and the Vietnam clade, have monomorphic fronds, the strongly dimorphic fronds in the Colysis clade and the L. ellipticus clade should be a derived character state. Based on character analyses, the strongly dimorphic fronds evolved independently at least three times in the Colysis clade and four times in the L. ellipticus clade. Within the Colysis clade, all species have strongly dimorphic fronds, except L. multilobus and L. hemionitideus (Fig. 8). The relationships of these latter two species with others in the clade remain unclear.

4.4.3. Rounded and elongate sori being plesiomorphic

Out of our currently recognized 51 species of Leptochilus (Zhang et al., 2024), most species have linear and acrostichoid sori, while only five species have rounded or elongate sori, including two species in the L. pteropus clade, one in the Colysis clade, and two in the L. ellipticus clade (Fig. 9). Therefore, the rounded and elongate sori in Leptochilus might have been derived from linear or acrostichoid sori. However, we argue that rounded and elongate sori is plesiomorphic instead, because: (1) the L. pteropus clade, the first-diverging lineage, has rounded or elongate sori; (2) genera closely related to Leptochilus, such as Phymatosorus, Microsorum sensu lato, also have rounded or elongate sori; and (3) intermediate forms of soral shapes are observed in some species, such as L. evrardii, L. hemionitideus, and L. macrophyllus, suggesting that linear or acrostichoid sori may actually be coenosori fused by rounded or elongate sori. Thus, the linear or acrostichoid sori in Leptochilus are different from those in Acrostichum L. (Pteridaceae), Blechnum L. (Blechnaceae) or Pteris L. (Pteridaceae), but may be analogous to Selliguea and Pleopeltis in Polypodiaceae (Christensen, 1929; Wagner, 1986; Otto et al., 2009).

4.5. Diversification of different habits in Leptochilus

The habit of Leptochilus species can vary from terrestrial to epipetric (epilithic) and even epiphytic (Zhang and Nooteboom, 2013), although detailed information about the habits of most species remains insufficiently documented. Including 16 species of Leptochilus, Chen et al. (2023) investigated the habitat preferences in Polypodiaceae and showed that the exclusively epiphytic habit of Leptochilus axillaris evolved once from the ancestors with terrestrial or rheophytic habitats. Recently, two new species with hemiepiphytic habitats, L. ornithopus and L. scandens, have been described (Fujiwara et al., 2023; Wei et al., 2023; Yu et al., 2024). Our field observations found that L. luangprabangensis and L. vietnamensis thrived on tree trunks on limestone mountains, contrasting the morphologically similar L. pedunculatus growing beside or on rocks in forest with basic soil. The adaptation to the epipetric and epiphytic habit may have played an important role in accelerating the diversification, but this needs further analysis. Among the 11 new species, L. bolikhamsaiensis, L. daklakensis, L. khammouanensis, L. locii, L. neolongipes, L. sinovietnamica, and L. wusugongii are epipetric on calcareous rocks; the sporophytes of L. kachinensis, L. vietnamensis, and L. luangprabangensis occur on tree trunks, while L. multilobus grows on basic cliffs. It is noteworthy that Leptochilus species may exhibit both hemiepiphytic and lithophytic habits. For instance, among ca. 1000 observed individuals of L. brevipes, approximately one-third were potentially hemiepiphytic, while two-thirds were lithophytic. Our recently published new species, L. kachinensis, L. luangprabangensis, and L. vietnamensis, were observed to be hemiepiphytic but can also be lithophytic pending further field investigations. Careful field examinations of the growth habits of both gametophytes and sporophytes are crucial for determining the exact habit in Leptochilus, contributing to a better understanding of speciation and diversification of the genus.

4.6. Indo-Burma as a biodiversity hotspot for ferns

The species diversities of various groups of organisms in the Indo-Burma Biodiversity Hotspot have been well documented. For example, van Dijk et al. (2004) listed 74 endemic bird species out of the 1277 species found in Indo-Burma, 71 endemic mammal species out of the 430 in the hotspot including the famous saola not known to science until in 1992, 189 endemic non-marine reptile species out of the 519 species including probably the highest diversity of freshwater turtles in the world, 139 endemic amphibian species out of the 323 species in the hotspot, and 566 endemic freshwater fish out of the remarkable 1262 documented species, accounting for about 10 percent of the world total. Although, van Dijk et al. (2004) also estimated the total plant diversity in the hotspot to be about 13, 500 vascular plant species, of which about 7000 (52%) are endemic, there was no details about the fern diversity in the hotspot.

Tordoff et al. (2012) listed 13 vascular plants and 9 animal species which they defined as trigger species occurring only in one key biodiversity area globally, none of which were ferns or lycophytes. In fact, our current study shows that 30 (59% of total 51) species of Leptochilus occur in this hotspot, 24 (80% of the 30 species) of which are endemic to the Indo-Burma hotspot (Table 2; Fig. 1). Our preliminary data showed that there might be 80–100 species in the genus, ca. 67% of the new members will be from the Indo-Burma hotspot (Zhang et al., unpubl. data).

Leptochilus is not the only fern genus with rich diversity and endemism in the Indo-Burma hotspot. Zhang et al. (2017) and Zhou and Zhang (2019) found that 10 out of the 24 species in Pteridrys C.Chr. & Ching (Pteridryaceae) are endemic to a single country, Vietnam, and that 14 (58%) out of the 24 total species in Pteridrys occur in the hotspot, 11 (79%) of which are endemic to the hotspot. There are 10 (17%) out of ca. 60 total species of Hymenasplenium Hayata (Aspleniaceae) occurring in the hotspot, all of which are endemic to the hotspot (Xu et al., 2018, Xu et al., 2019a, Xu et al., 2019b, Xu et al., 2023). Ten out of the 22 species of Pleocnemia C. Presl (Dryopteridaceae) are distributed in the hotspot, five of which are endemic there (Xu and Zhang, unpubl. data). Five out of the total seven species of Rhachidosorus Ching (Rhachidosoraceae) occur in the hotspot and all of them are endemic in the hotspot (He and Kato, 2013). All three species of Diplaziopsis C.Chr. occur in the hotspot, though none of them are endemic. The only species of Brainea J.Sm. is also distributed in this hotspot.

Notably, the Indo-Burma hotspot was not even recognized by Suissa et al. (2021) as one of the eight hotspots they identified for fern diversity. Unfortunately, the fern diversity and its endemism in the Indo-Burma hotspot have not been well understood in spite of the online Ferns of Thailand, Laos and Cambodia (Lindsay and Middleton, 2012 onwards). Extensive explorations into the fern diversity in Cambodia, Laos, Myanmar, and Vietnam and a comprehensive analysis are desperately needed. From what we understood now, the Indo-Burma hotspot is definitely a diversity hotspot for ferns.

Acknowledgments

The research was supported by the Second Tibetan Plateau Scientific Expedition and Research (STEP) program (2019QZKK0502), a grant from Yunnan Fundamental Research Projects (Grant # 202201BC070001), a Yunnan Revitalization Talent Support Program "Young Talent" Project and a CAS Scholarship to Liang Zhang. We thank Daniele Cicuzza, En-De Liu, and Shi-Wei Yao for sharing DNA materials.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pld.2024.08.005.