Introduction

Leucobryoid mosses are haplolepideous mosses (Dicranidae) whose gametophytes exhibit a characteristic whitish green color. This color is caused by a leaf anatomy with a very wide costa composed of two to several layers of large, hyaline cells interconnected by pores (called leucocysts or hyalocysts), embedding 1–3 layers of chlorophyllose cells (chlorocysts).

Schimper (1856) recognized four leucobryoid moss genera (Arthrocormus, Leucobryum, Leucophanes and Octoblepharum) in a single family, the Leucobryaceae. The leucobryoid genera Cladopodanthus and Schistomitrium were not mentioned by Schimper (1856), and three further genera (Cardotia, Exodictyon and Ochrobryum) were described later during the nineteenth century. All these genera share a generally similar habit and leaf structure, but differ in their leaf anatomy (Cardot 1899; Fleischer 1904) and peristome characters (e.g., the presence of a preperistome, thickening and ornamentation of inner and outer plates). Based on these differences, Fleischer (1904) separated the leucobryoid genera into two families: the Leucobryaceae, considered more related to the Dicranaceae, and the Leucophanaceae, related to the Calymperaceae. Several other classifications for the aforementioned genera as well as further leucobryoid genera described during the twentieth century (Carinafolium, Exostratum, Holomitriopsis and Steyermarkiella) were proposed (overview in Yamaguchi 1993). The latest classification of mosses previous to the introduction of molecular phylogenetic studies (Buck and Goffinet 2000) recognized 11 leucobryoid genera, classified in the Dicranaceae (Holomitriopsis and Steyermarkiella), the Calymperaceae (Arthrocormus, Exodictyon, Exostratum, Leucophanes, and Octoblepharum) and the Leucobryaceae (Cladopodanthus, Leucobryum, Ochrobryum, and Schistomitrium).

That the leucobryoid genera in the Calymperaceae (and the Octoblepharaceae, cf. Bonfim Santos and Stech 2017) were not closely related to those in the Leucobryaceae, as proposed by Fleischer (1904), was confirmed by molecular data (e.g., Bonfim Santos and Stech 2017; Cox et al. 2010; Fisher et al. 2007; La Farge et al. 2000). In contrast, molecular phylogenetic evidence suggested that a number of the Dicranaceae genera belong to the Leucobryaceae (Hedderson et al. 2004; La Farge et al. 2000, 2002; Stech 1999, 2004; Tsubota et al. 2003, 2004). These genera display morphologies that are either somewhat leucobryoid (Brothera, Campylopodiella) or dicranoid (e.g., Atractylocarpus, Campylopus, Dicranodontium, Pilopogon). The typical dicranoid costa has a median band of enlarged deuter cells surrounded by dorsal and ventral layers of stereids, and dorsal and ventral epidermal layers (Frahm 1991). This basic structure, however, varies considerably in different genera, e.g., by the absence of epidermal or stereid layers or their further subdivision into more layers. The so-called paraleucobryoid costa, named after the genus Paraleucobryum (Dicranaceae), is characterized by median chlorocysts (corresponding to the deuter cells) surrounded by ventral and dorsal hyalocysts, thus resembling the costa of Leucobryum. The term paraleucobryoid has also been applied to the costa of Brothera and Campylopodiella, which, however, differs by the presence of stereids. In contrast to the hyalocysts of the leucobryoid costa, the hyalocysts in Paraleucobryum, Brothera and Campylopodiella lack interconnecting pores between cells and never form multiple layers.

In the classification of mosses by Goffinet and Buck (2004) and Goffinet et al. (2009), seven dicranoid genera (Atractylocarpus, Brothera, Bryohumbertia, Campylopodiella, Campylopus, Dicranodontium and Microcampylopus) were included in the Leucobryaceae. The classification in Frey and Stech (2009) differed from Goffinet and Buck (2004) and Goffinet et al. (2009) by placing Microcampylopus in the Dicranellaceae, adding Holomitriopsis, Mitrobryum, Sphaerothecium and Steyermarkiella (Dicranaceae in Goffinet et al. 2009) to the Leucobryaceae, and incorporating nomenclatural changes at genus level, synonymizing Bryohumbertia with Campylopus (Stech 2004) and mistakenly adopting the use of Atractylocarpus for the Campylopodiella species and of Metzleria for the Atractylocarpus species (a misinterpretation of the nomenclatural proposal by Frahm and Isoviita 1988). The placement of Microcampylopus in the Dicranellaceae and of Holomitriopsis in the Leucobryaceae was supported by molecular data (Bonfim Santos and Stech 2017; Cox et al. 2010; Stech 2004; Stech et al. 2012). Mitrobryum, Sphaerothecium and Steyermarkiella have not yet been included in molecular phylogenetic studies.

While the Leucobryaceae are molecularly well circumscribed, molecular data to assess generic delimitations and relationships within the family are still limited. Cox et al. (2010) covered 10 out of 14 genera of the Leucobryaceae sensu Frey and Stech (2009), which were resolved in three lineages: a dicranoid lineage comprising Campylopus and Pilopogon (the sequences named as Microcampylopus leucogaster in the same clade originate from a collection that actually belongs to Campylopus); another dicranoid lineage comprising Brothera, Campylopodiella and Dicranodontium; and a leucobryoid lineage with Ochrobryum, Holomitriopsis, Leucobryum, Cladopodanthus and Schistomitrium. A similar topology including representatives of eight genera was resolved in Stech et al. (2012). However, relationships between the major lineages within the Leucobryaceae were contradictory between these and other molecular studies, and the taxon sampling was too limited to infer generic circumscriptions.

Furthermore, uniting gametophytically heterogeneous genera in the Leucobryaceae considerably obscured the family’s morphological circumscription. The evolution of morphological traits within the Leucobryaceae and the suitability of morphological characters to assess generic delimitations and relationships remain insufficiently known, since all revisional work concerning these genera predates molecular phylogenetic studies.

Consequently, the present study aims to (1) test hypotheses of suprageneric relationships and genus circumscriptions in the Leucobryaceae based on phylogenetic analyses of a comprehensive taxon and molecular marker sampling and (2) infer the evolution of morphological characters within the family based on ancestral state reconstructions.

Materials and methods

Taxon sampling, DNA extraction and sequencing

The taxon sampling comprised 63 Leucobryaceae specimens (representing 11 out of 14 genera and 45 species) as well as Archidium alternifolium (Archidiaceae), Eustichia longirostris (Eustichiaceae) and Micromitrium tenerum (Micromitriaceae) as outgroup representatives following earlier phylogenetic reconstructions (Fedosov et al. 2016; Goffinet et al. 2011; Stech et al. 2012). Molecular markers from all three genomes were sequenced: mitochondrial (mt) nad5 G1 intron, chloroplast (cp) trnS-trnF region and atpB-rbcL spacer and nuclear (nr) ribosomal ITS1-5.8S-ITS2 (ITS) region. Sequences were obtained from GenBank and newly sequenced specimens from herbaria L, MO, SING and UB. Voucher information and GenBank accession numbers are listed in Online Resource 1.

Procedures for DNA extraction, amplification and sequencing followed Bonfim Santos and Stech (2017). Sequences were manually aligned in Geneious® v8.0.5 (Biomatters Ltd.), using the alignment from Stech et al. (2012) as a starting point.

Phylogenetic reconstructions

Three alignments were analyzed. The first alignment comprised the combined mitochondrial and chloroplast markers for all Leucobryaceae specimens, with Archidium alternifolium, Eustichia longirostris and Micromitrium tenerum as outgroup representatives. The second and third alignments represented two major clades within the Leucobryaceae, the Dicranodontium and the leucobryoid clades, respectively, and included additionally ITS, which was in parts unalignable across the whole family due to high sequence variability. Two samples of Ochrobryum gardneri were used as outgroup representatives, based on the analyses of the first alignment and alignability of the ITS sequences. The third major lineage within the Leucobryaceae, the Campylopus clade, has been extensively studied elsewhere (Stech 2004; Stech and Dohrmann 2004; Stech and Wagner 2005; Stech et al. 2010) and therefore was not studied in detail in this work. Alignment lengths as well as numbers of variable and parsimony-informative positions are provided in Online Resource 2.

Phylogenetic reconstructions were performed under maximum likelihood (ML) using RAxML v.8 (Stamatakis 2014) and Bayesian inference (BI) using MrBayes v.3.2.6 (Ronquist et al. 2012), both on the CIPRES Science Gateway v.3.3 (Miller et al. 2010). Analyses were run for the markers separated per genome (mt, cp, nr) and for the complete alignments (mt + cp or mt + cp + nr, respectively), to check for incongruence and to infer how each genome contributed to the resolution and clade support. Gaps were treated as missing data. Selection of partitioning schemes and evolutionary model testing were performed in PartitionFinder v1.1.1 (Lanfear et al. 2012) for the models that can be implemented in RAxML (GTR) and MrBayes (GTR and several of its nested models), respectively, with or without a gamma-distributed rate variation among sites (Γ) and a proportion of invariable sites (I). Model parameters were estimated independently for each partition. The resulting best partitioning schemes and evolutionary models according to the Akaike information criterion (AIC) were implemented in the ML and BI analyses (Online Resource 3). In RAxML, a single type of rate heterogeneity pattern (either + Γ, + I or + Γ + I) can be applied for all partitions per analysis; thus, we implemented GTR + Γ in all ML analyses. For all maximum likelihood analyses, rapid bootstrapping with 1000 iterations was performed. For Bayesian inference, four runs with four chains (5 × 106 generations each) were run simultaneously, with the temperature of the single heated chain set to 0.4. Chains were sampled every 1000 generations, and the respective trees were written to tree files. After verifying the convergence of runs in Tracer v1.6 (Rambaut et al. 2014), fifty percent majority rule consensus trees and PP of clades were calculated, discarding the burn-in phase (25%).

The Shimodaira–Hasegawa (SH) test (Goldman et al. 2000; Shimodaira and Hasegawa 1999) was applied to compare selected alternative phylogenetic hypotheses for the Leucobryaceae and the Dicranodontium clade. For the Leucobryaceae, the ML tree (topology as in Fig. 1) was compared with the hypotheses of (1) the sister group relationship of the Campylopus clade and the Dicranodontium clade (as resolved in Stech 2004), (2) the sister group relationship of the Campylopus clade and the leucobryoid clade (Tsubota et al. 2004), (3) a monophyletic Leucobryum (as resolved in the analyses of the leucobryoid clade alignment) and (4) a clade formed by Cladopodanthus, Holomitriopsis, Ochrobryum and Schistomitrium, a relationship suggested by Eddy (1990) and Robinson (1990). For the Dicranodontium clade, the ML tree (topology as in Fig. 2) was compared with the hypotheses of (1) a monophyletic Campylopodiella, (2) a monophyletic Dicranodontium, (3) a monophyletic Dicranodontium including D. subporodictyon, (4) a monophyletic Dicranodontium including Atractylocarpus intermedius and (5) a monophyletic Dicranodontium including both D. subporodictyon and A. intermedius, all hypotheses based on the taxonomic literature (Allen 1992a; Allen and Ireland 2002; Frahm 1997; Müller and Frahm 1987). Constraint trees were used as an input to ML with RAxML. The resulting trees with branch length values and corresponding alignment were loaded into PAUP* v.4.0b10 (Swofford 2002), where these trees were compared with the respective unconstrained topologies using the SH test with 10,000 bootstrap replicates and the resampling estimated log-likelihood method (RELL).

Fig. 1
figure 1

Bayesian inference consensus tree of 63 Leucobryaceae representatives based on mitochondrial and chloroplast DNA sequences (nad5, trnS-trnF region and atpB-rbcL). Archidium alternifolium (Archidiaceae), Eustichia longirostris (Eustichiaceae) and Micromitrium tenerum (Micromitriaceae) were used as outgroup representatives. Branch support is indicated for Bayesian inference (BI) and maximum likelihood (ML) analyses of the same alignment. Bold branches represent posterior probabilities (PP) ≥ 0.95. Actual bootstrap (BS) values are shown if ≥ 70%

Full size image
Fig. 2
figure 2

Bayesian inference consensus tree of 13 representatives of the Dicranodontium clade of the Leucobryaceae based on nuclear, mitochondrial and chloroplast DNA sequences (ITS, nad5, trnS-trnF region and atpB-rbcL). Two samples of Ochrobryum gardneri (Leucobryaceae) were used as outgroup representatives. Branch support is indicated for Bayesian inference (BI) and maximum likelihood (ML) analyses of the same alignment. Bold branches represent posterior probabilities (PP) ≥ 0.95. Actual bootstrap (BS) values are shown if ≥ 70%

Full size image

The NeighborNet algorithm (Bryant and Moulton 2004) implemented in SplitsTree4 (Huson and Bryant 2006) was applied to the Leucobryaceae alignment for visualization of the data in a phylogenetic network. Missing data obscured the network patterns; thus, all specimens with missing data for an entire molecular marker were removed from the alignment.

Ancestral state reconstructions

Maximum likelihood ancestral state reconstructions were performed with Mesquite v.3.2 (Maddison and Maddison 2017) under the Markov k-state model (Lewis 2001). The analyzed characters for the Leucobryaceae were leucobryoid morphology (0 absent, 1 present), seta orientation when young or moist (0 straight, 1 twisted, 2 cygneous), capsule orientation (0 orthotropous, 1 homotropous) and calyptra shape (0 cucullate, 1 mitrate, 2 reduced). The analyzed characters for the Dicranodontium clade were the ventral costa layer (0 differentiated in ventral epidermis and stereid band below, 1 stereid band, 2 hyalocysts), dorsal costal stereids (0 in groups, 1 in a continuous band, 2 absent), cell type in the dorsal epidermal layer of costa (0 chlorocysts, 1 stereids, 2 hyalocysts), occurrence of pitted basal lamina cells (0 absent, 1 present) and seta orientation when young or moist (0 straight, 1 twisted, 2 cygneous). For both alignments, the character evolution analyses were performed on the ML tree and on the constrained ML trees representing plausible alternative hypotheses according to the SH test results (Leucobryaceae: sister group relationship of the Campylopus clade and the Dicranodontium clade, sister group relationship of the Campylopus clade and the leucobryoid clade, and Leucobryum monophyletic; Dicranodontium clade: Campylopodiella and Dicranodontium reciprocally monophyletic).

Results

Phylogenetic reconstructions

Figures 1, 2 and 3 show consensus trees from Bayesian inference of the concatenated mitochondrial and chloroplast markers for the Leucobryaceae, all concatenated markers for the Dicranodontium clade and all concatenated markers for the leucobryoid clade, respectively. Branch support values are indicated for Bayesian inference (posterior probabilities, PP) and maximum likelihood analyses (bootstrap support, BS). Trees resulting from analyses of subsets of data are shown in Online Resource 4a–p, and relevant aspects of their topologies and branch support patterns are mentioned below. The trees obtained from analyses of markers from each genome separately differed in resolution and branch support, but did not reveal statistically supported incongruence (i.e., no PP ≥ 0.95 or BS ≥ 70% for the conflicting branches in both incongruent topologies). They were also congruent with the trees resulting from the combined analyses, except for the BI analyses of the leucobryoid clade (see below).

Fig. 3
figure 3

Bayesian inference consensus tree of 36 representatives of the leucobryoid clade of the Leucobryaceae based on nuclear, mitochondrial and chloroplast DNA sequences (ITS, nad5, trnS-trnF region and atpB-rbcL). Two samples of Ochrobryum gardneri (Leucobryaceae) were used as outgroup representatives. Branch support is indicated for Bayesian inference (BI) and maximum likelihood (ML) analyses of the same alignment. Bold branches represent posterior probabilities (PP) ≥ 0.95. Actual bootstrap (BS) values are shown if ≥ 70%

Full size image

The Leucobryaceae genera included in this study were split in three well-supported clades (Fig. 1), the Campylopus clade (PP 1, BS 100%), which comprised the dicranoid genera Campylopus (PP 1, BS 100%) and Pilopogon (PP 1, BS 98%), the Dicranodontium clade (PP 1, BS 100%), which comprised the remaining dicranoid genera (Atractylocarpus, Brothera, Campylopodiella, Dicranodontium) and the leucobryoid clade (PP 1, BS 99%), which comprised the leucobryoid genera (Cladopodanthus, Holomitriopsis, Leucobryum, Ochrobryum, Schistomitrium). Within the leucobryoid clade, Cladopodanthus and Schistomitrium were both monophyletic and formed sister clades with maximum support (PP 1, BS 100%). They appeared in these analyses as sister to Leucobryum sanctum (PP 1, BS 82%). The Leucobryum/Cladopodanthus/Schistomitrium clade was monophyletic (PP 1, BS 100%) and sister to Holomitriopsis (PP 1, BS 95%). Ochrobryum was monophyletic (PP 1, BS 100%) and sister to the clade formed by all other leucobryoid genera. The sister group relationship of the Dicranodontium clade and the leucobryoid clade was not significantly supported by BI, but well supported by ML (BS 92%), while in the analyses of the chloroplast markers separately it was supported by both methods (PP 0.99, BS 91%). Analyses of the chloroplast markers resulted in trees with the same topology and similar branch support as the analyses of all markers combined, whereas in analyses of nad5 alone relationships within the main Leucobryaceae clades were unresolved or weakly supported (Online Resource 4a–d).

Within the Dicranodontium clade (Fig. 2), Atractylocarpus was monophyletic (PP 0.97, BS 82%) and the two Brothera leana samples formed a clade with maximum support (PP 1, BS 100%), whereas Campylopodiella was not monophyletic, with a well-supported sister group relationship between Brothera leana and Campylopodiella flagellacea (PP 1, BS 95%). Dicranodontium was not monophyletic either, since the D. pulchroalare/D. porodictyon clade (PP 1, BS 100%) was sister to the weakly supported Brothera/Campylopodiella clade (PP 0.96). Dicranodontium subporodictyon Broth. was well supported within this clade as sister to the Atractylocarpus clade (PP 0.97, BS 80%). Resolution and branch support were highest in the separate ITS analyses, followed by the cp markers, and nad5 provided the least resolution (Online Resource 4e–j).

For the leucobryoid clade, the results shown in Fig. 3 mostly agreed with those in Fig. 1, except for the relationships between the Cladopodanthus/Schistomitrium clade and Leucobryum species. In the analyses of the leucobryoid clade (Fig. 3), the Cladopodanthus/Schistomitrium clade (PP 1, BS 100%) and the monophyletic Leucobryum (PP 0.99) were sister groups (PP 1, BS 95%), while as shown in Fig. 1, Cladopodanthus and Schistomitrium were nested within Leucobryum, causing Leucobryum to be paraphyletic. Bayesian inference analyses of the leucobryoid clade for each genome did not resolve the same relationships as all markers combined, but repeated the topology as seen in Fig. 1. Maximum likelihood analyses for all markers combined recovered a monophyletic Leucobryum as well, but with low support, and thus did not represent an incongruence in relation to the analyses per genome. Chloroplast markers were the most informative for the relationships within the leucobryoid clade, followed by nad5, and ITS provided the least resolution (Online Resource 4k–p).

The SH test applied to the Leucobryaceae alignment (Table 1) did not reject the two alternative hypotheses of sister group relationships between the three main Leucobryaceae clades, nor the hypothesis of Leucobryum being monophyletic (as in Fig. 3) as significantly less likely than the unconstrained ML topology (as in Fig. 1). The alternative topology with Cladopodanthus, Holomitriopsis, Ochrobryum and Schistomitrium forming one clade was rejected. The SH test applied to the Dicranodontium clade alignment (Table 1) did not reject the hypotheses of the monophyly of Campylopodiella and of Dicranodontium, but the hypotheses of a broader circumscription of Dicranodontium, including D. subporodictyon and/or Atractylocarpus intermedius, were rejected.

Table 1 Results from the SH tests applied to the Leucobryaceae alignment and to the Dicranodontium clade alignment
Full size table

As a graphic representation of the distances between the aligned sequences, the phylogenetic network for the Leucobryaceae alignment (Fig. 6) shows alternative relationships between Leucobryaceae representatives are possible than those resolved by our analyses. By far, most of the sequence divergence represented in the graph is found after the early splits between the three main Leucobryaceae clades, while distances in this initial evolution of the family are quite small and allow for the three alternative topologies of sister group relationships. Additionally, the network also shows the uncertainty regarding the relationships between representatives of Leucobryum and the genera Cladopodanthus and Schistomitrium.

Ancestral state reconstructions

Ancestral state reconstructions of the Leucobryaceae resolved the leucobryoid morphology as a derived character which originated in the most recent common ancestor (MRCA) of the leucobryoid clade (Fig. 4a). The cygneous seta originated at least twice, in the MRCA of the genus Campylopus and within the Dicranodontium clade (Fig. 4b). According to this analysis, the twisted seta originated in the MRCA of the Dicranodontium clade. However, due to the low resolution of the relationships within the Dicranodontium clade in the analyses for the entire family (Fig. 1), character evolution in this clade is better interpreted in the analyses for the Dicranodontium clade (Figs. 2, 5). Homotropous capsules originated twice, in the MRCAs of Campylopus and Leucobryum, and reversed to the plesiomorphic state in the MRCA of Cladopodanthus and Schistomitrium (Fig. 4c). The mitrate calyptra originated twice, in the MRCAs of Ochrobryum and Cladopodanthus/Schistomitrium (Fig. 4d). The reconstructions under the alternative hypotheses differed only for the capsule orientation under the hypothesis of a monophyletic Leucobryum, which would have originated twice, in the MRCAs of Campylopus and Leucobryum, without reversals to the plesiomorphic state (Online Resource 5).

Fig. 4
figure 4

Maximum likelihood character evolution analyses for the occurrence of leucobryoid morphology (a), seta orientation (b), capsule orientation (c) and calyptra shape (d) for the Leucobryaceae alignment, under the phylogenetic hypothesis represented by the unconstrained maximum likelihood tree

Full size image
Fig. 5
figure 5

Maximum likelihood character evolution analyses for the ventral costa layer (a, f), dorsal stereids (b, g), cell type in the dorsal costa epidermis (c, h), occurrence of pitted basal lamina cells (d, i) and seta orientation (e, j) for the Dicranodontium clade alignment, under two different phylogenetic hypotheses: the unconstrained maximum likelihood tree (a–e, left) and the constrained maximum likelihood tree with Campylopodiella and Dicranodontium monophyletic (f–j, right)

Full size image

For the Dicranodontium clade, the analysis based on the ML tree (Fig. 5a–e, left) resolved the ventral costa layers forming a stereid band as originating independently within A. intermedius and in the ancestors of C. flagellacea and C. stenocarpa. The hyalocyst layer in the ventral costa originated in the MRCA of Brothera (Fig. 5a). The isolated stereid groups originated either twice, in the MRCA of Brothera/Campylopodiella flagellacea and in the ancestral of C. stenocarpa, or less likely only once, in the previous node (Fig. 5b). The stereids forming the dorsal epidermis originated in the MRCA of Atractylocarpus/D. subporodictyon, while hyalocysts originated either twice, in the MRCA of Brothera/Campylopodiella flagellacea and in the ancestor of C. stenocarpa, or less likely only once in the previous node (Fig. 5c). Pitted basal lamina cells most likely originated in the MRCA of Atractylocarpus/D. subporodictyon (Fig. 5d). The cygneous seta originated in the MRCA of the entire Dicranodontium clade and was modified to an erect and twisted seta in the Campylopodiella/Brothera clade and in the MRCA of A. longisetus/A. alticaulis (Fig. 5e). The analyses based on the hypothesis of both Campylopodiella and Dicranodontium monophyletic (Fig. 5f–j, right) differed in that the dorsal stereids in isolated groups would have originated in the MRCA of Brothera/Campylopodiella (Fig. 5g), as the hyalocysts in the dorsal epidermis (Fig. 5h) and the twisted seta (Fig. 5j).

Discussion

In line with earlier phylogenetic studies (Cox et al. 2010; Fedosov et al. 2015, 2016; Hedderson et al. 2004; La Farge et al. 2000, 2002; Stech 2004; Tsubota et al. 2003, 2004), the analyses of the present dataset resolved three well-supported lineages within the Leucobryaceae: the dicranoid Campylopus clade, the dicranoid Dicranodontium clade and the leucobryoid clade. The latter is characterized by the leucobryoid costa as synapomorphic character (Fig. 4a). The twisted seta could represent a synapomorphy for the Dicranodontium clade in the ancestral state reconstruction of the Leucobryaceae (Fig. 4b). However, Atractylocarpus intermedius and Dicranodontium present a cygneous seta, which more likely represents the ancestral character state in the analyses of the Dicranodontium clade separately (Fig. 5e, j). The evolution of a cygneous seta occurred twice in the Leucobryaceae, possibly from different ancestral states, namely from a twisted seta in Dicranodontium and from a straight seta in Campylopus (Fig. 4b). The other morphological characters analyzed here (capsule orientation and calyptra shape) changed character states within the Campylopus and/or leucobryoid clades, and thus do not represent synapomorphies for either clade. Contrary to what was suggested by Robinson (1990), Leucobryum is a derived genus and its homotropous capsules are not a plesiomorphic trait shared with Campylopus, but evolved independently in both genera (Fig. 4c). Thus, the orthotropous capsules of the remaining leucobryoid genera do not represent evidence of shared ancestry, and neither does the mitrate calyptra found in Ochrobryum and in the Cladopodanthus/Schistomitrium clade (Fig. 4d).

Earlier phylogenetic analyses could not resolve the relationships between the three major Leucobryaceae clades (Hedderson et al. 2004) or supported different sister group relationships: Campylopus clade and leucobryoid clade (Tsubota et al. 2004), Campylopus clade and Dicranodontium clade (one tree in Stech 2004) and, most frequently, Dicranodontium clade and leucobryoid clade (Cox et al. 2010; Fedosov et al. 2015, 2016; La Farge et al. 2000, 2002; Stech et al. 2012; Tsubota et al. 2003, 2004; and another tree in Stech 2004). Three studies (Cox et al. 2010; La Farge et al. 2002; Tsubota et al. 2003, 2004) recovered the latter topology with ≥ 70% bootstrap support or ≥ 0.95 Bayesian posterior probability. Our results corroborate these studies, since our analyses also recovered the sister group relationship of the Dicranodontium and leucobryoid clades, supported in ML analyses for the total alignment (Fig. 1) and in both BI and ML analyses for the chloroplast markers (Online Resource 4c, d). However, the three possible alternative topologies of sister group relationships could not be rejected based on the SH test performed on our data (Table 1, Leucobryaceae alignment). Relationships between the major lineages within the Leucobryaceae thus remain somewhat uncertain even with a larger taxon and marker sampling than in previous studies. The patterns observed in the phylogenetic network (Fig. 6) indicate that the early evolution of the Leucobryaceae may have been an event of rapid radiation, with the least sequence divergence occurring until the split of the three main Leucobryaceae lineages. The shorter the branches, e.g., the least substitutions the branches represent, the harder it is to reconstruct the relationships associated with them, and this phylogenetic network puts in evidence this challenge in the phylogenetic reconstruction for the early evolution of the Leucobryaceae.

Fig. 6
figure 6

NeighborNet phylogenetic network of 30 Leucobryaceae representatives based on nuclear, mitochondrial and chloroplast DNA sequences (ITS, nad5, trnS-trnF region and atpB-rbcL). The outgroup representatives Archidium alternifolium (Archidiaceae) and Eustichia longirostris (Eustichiaceae) are included in the graph

Full size image

We found the Dicranodontium clade to have the lowest sequence variability for the molecular markers applied in this study (except for ITS), resulting in short branch lengths in the phylogenetic reconstructions (Figs. 1, 3). This clade is also the least species rich (ca. 30 species, Frey and Stech 2009). Its species have the narrowest habitat range: temperate to subtropical and montane/alpine tropical regions, mostly in moist forest habitats (except some Atractylocarpus only found above the tree line) and absent from tropical lowlands (Frahm 1991; Padberg and Frahm 1985), and the least variation in costa structure. The latter ranges from a typically dicranoid costa in Dicranodontium to variations with reduced numbers of cell types or modifications on the ventral side (hyalocysts in Brothera, stereids in Campylopodiella, Fig. 5a, f) and the dorsal side (stereids in Atractylocarpus, hyalocysts in Brothera/Campylopodiella, Fig. 5c, h). Dorsal stereids remain present, albeit partly in reduced numbers, in all genera, either in groups or as a continuous stereid band (Fig. 5b, g).

The Campylopus and leucobryoid clades are similar in containing more molecular variation than the Dicranodontium clade, indicated by their long branch lengths (Figs. 1, 3). The largest number of Leucobryaceae species (ca. 160, Frey and Stech 2009), the broadest distribution and habitat range (from latitude 70°N to 65°S, from sea level to 4800 m a.s.l.; Frahm 1991) and the greatest variety of modifications of the dicranoid costa structure are found in the Campylopus clade. Campylopus and Pilopogon species may either have the basic dicranoid costa structure, or be modified in various ways. Part of the species have a ventral epidermis of chlorocysts or stereids, with the dorsal epidermis ribbed (with protruding cells) or forming lamellae up to seven cells high (e.g., Gama et al. 2016), or with all costa layers but the deuter cells reduced to stereids. Other Campylopus species have ventral hyalocysts that may cover more than half of the costa section, with the dorsal epidermis consisting of smaller cells (smooth, ribbed or forming lamellae) or also of hyalocysts (Frahm 1983, 1991). The leucobryoid clade is also species rich (ca. 110 species, Frey and Stech 2009). Its species are concentrated in the tropics (with the exception of some Leucobryum species), with maximum diversity in tropical and subtropical rainforests (Eddy 1990). The costa in this clade is highly modified, although rather invariable when compared to the Campylopus clade. Its leucobryoid pattern varies solely in the number of ventral and dorsal hyalocyst layers.

The findings discussed above indicate that patterns of molecular variation, species richness, geographical distribution, ecological amplitude and of variation in costa structure covary in the three main lineages of the Leucobryaceae. The two lineages with the most modified morphologies are also the most molecularly variable, species diverse, and occupy the broadest distributions and widest variety of habitats. Thus, it can be hypothesized that costa structure modifications, by allowing an improved exploitation of the available ecological niches and environment resources, could have triggered higher phylogenetic diversity in the Campylopus and leucobryoid clades. Within the Campylopus clade, the variety of modified costa forms may be a response to (or perhaps the cause of) the broad ecological spectrum of the genus and may represent different optimization strategies for photosynthesis, water uptake, water storage and mechanical fixation (Frahm 1985). The rather invariable leucobryoid costa, in contrast, seems to be most successful in quite distinct environments, possibly representing a strategy to optimize gas exchange and water balance in the damp habitat of tropical forests (Robinson 1985, 1990). It is not, however, restricted to the Leucobryaceae, but appears as a derived state in at least two other families of haplolepideous mosses, the Calymperaceae (genera Arthrocormus, Exodictyon, Exostratum and Leucophanes) and the Octoblepharaceae (Octoblepharum) (Bonfim Santos and Stech 2017; Cox et al. 2010; Fisher et al. 2007; La Farge et al. 2000), which occur mainly in tropical rainforests as well.

An evolutionary history with species-rich long branch clades and short branch clades with much lower species diversity was also observed in the flowering plant family Annonaceae (Richardson et al. 2004). However, later studies of the Annonaceae have shown that differences in species numbers could not be attributed to diversification rate shifts, nor could the observed rate shifts be correlated with key morphological innovations (Erkens et al. 2012). Whether this is the case also in the Leucobryaceae remains to be tested.

The presently estimated phylogenetic relationships raise doubts concerning the delimitation of some of the genera in the Dicranodontium clade (Fig. 2) and about the monophyly of Leucobryum (Figs. 1, 3). Within the Dicranodontium clade, the genus Atractylocarpus is molecularly well supported (this study) and distinguished from the other genera by its leaves gradually tapering into a long subula, the position of rhizoid initials and pitted basal lamina cells (Frahm 1991, Padberg and Frahm 1985; Fig. 5d, i). According to our results, its circumscription should include Dicranodontium subporodictyon. This species received much attention due to its peculiar disjunct distribution pattern (British Columbia/Canada, Scotland/UK, Sikkim/India and Yunnan Province/China). Its systematic position remained unclear because of incompatible gametophytic characters with each of the three genera in which it was placed, i.e., Campylopus, Dicranodontium (Leucobryaceae) and Dicranum (Dicranaceae), aggravated by the fact that its sporophytes are still unknown (Allen and Ireland 2002; Chien and Tong 1992; Corley and Wallace 1974; Frahm 1997). The present molecular data unequivocally support a placement of D. subporodictyon in the Dicranodontium clade, where it appears as sister to the Atractylocarpus species included in our study. This relationship could have been predicted since the species displays the diagnostic characters of Atractylocarpus (long subulate leaves, incrassate, pitted lamina cells and position of rhizoid initials), which seem to have been overlooked or misinterpreted in previous studies. Consequently, we propose a new combination here (see Taxonomic treatment). The monospecific Brothera can be recognized by the absence of ventral stereids and the presence of hyalocysts in its costa (Frahm 1991; Müller and Frahm 1987).

The delimitations of Campylopodiella and Dicranodontium, however, are less clear. Both genera were resolved as paraphyletic (Fig. 2), although the SH test results did not reject the hypotheses of their monophyly (see Table 1). As far as morphological characters are concerned, the cygneous seta, as discussed above, does not represent a synapomorphy for the genus Dicranodontium, but also occurs in Atractylocarpus, while the twisted seta is shared by Atractylocarpus and Campylopodiella (Fig. 5e, j), and elongate upper lamina cells occur in all genera of the Dicranodontium clade (Müller and Frahm 1987; Padberg and Frahm 1985). Dicranodontium, however, is the only genus of this clade with a typical dicranoid costa. Considering the uncertainty regarding generic limits in the Dicranodontium clade, we do not yet propose major changes in the classification. In case future studies support the paraphyly of Campylopodiella and Dicranodontium, a broader circumscription of a morphologically variable Dicranodontium may be adopted, which would be separated from Atractylocarpus by the characters listed above. Although Brothera and Campylopodiella have distinctive morphological characters in relation to Dicranodontium, those could be interpreted as “budding” diversification (Vanderpoorten and Long 2006).

In the leucobryoid clade, all genera but Leucobryum share a costa mostly formed by two hyalocyst layers at leaf base, hypocentric chlorocysts (closer to the dorsal surface in transverse section, due to a difference in the depth of the ventral and dorsal hyalocyst layers), capsules that are orthotropous and symmetrical, entire peristome teeth (except Ochrobryum, peristome absent) and a mitrate calyptra (except Holomitriopsis, cucullate) (Allen 1992b; Eddy 1990; Magill 1993; Robinson 1965). Leucobryum, in contrast, is characterized by irregularly subdivided hyalocysts forming three to several layers at some portions of the costa at leaf base, asymmetrical, homotropous to orthogonal, curved and gibbose capsules, peristome teeth divided to the middle and a cucullate calyptra (Yamaguchi 1993). Leucobryum subobtusifolium (Broth.) B.H.Allen, a species which was originally placed in Ochrobryum, but transferred to Leucobryum based on the presence of apical clusters of brood leaves instead of globose propagules (its sporophyte is unknown) (Allen 1992b), indeed belongs to the latter genus according to the present molecular analyses.

The conclusion by Eddy (1990) and Robinson (1990) that Cladopodanthus, Holomitriopsis, Ochrobryum and Schistomitrium are closely related and should be separated from Leucobryum in the family Schistomitriaceae A. Eddy (1990), however, is rejected by the present molecular data (Figs. 1, 3, Table 1). Our results suggest that the shared character states of these genera are either convergences (as the mitrate calyptra) or retained ancestral character states (as the orthotropous capsules), while at least part of the distinguishing traits of Leucobryum correspond to apomorphic character states within the leucobryoid clade.

The close relationship between Cladopodanthus/Schistomitrium and Leucobryum supported here is in agreement with results of previous studies (Cox et al. 2010; La Farge et al. 2000; Tsubota et al. 2004). However, our study provided conflicting results regarding the relationships between these genera. While the family-level analyses with nad5 and chloroplast markers support the sister group relationship of Cladopodanthus/Schistomitrium and the Asian species L. sanctum, and thus resolve Leucobryum as paraphyletic, the clade-level analyses (with additionally ITS) recover either this same topology (separate analyses of markers per genome, supported for mitochondrial and chloroplast markers) or the monophyly of Leucobryum (for all markers combined, Fig. 3, supported for BI only). The results of the SH test (Table 1) show that the available data do not support a preference for either hypothesis. Since morphology indicates that Leucobryum is monophyletic, the contradictory topology could be caused by plesiomorphic molecular characters shared by Leucobryum species (L. bowringii, L. crispum, L. giganteum and L. sanctum) as well as Cladopodanthus and Schistomitrium. On the other hand, the phylogenetic network (Fig. 6) puts in evidence the uncertainty of these relationships and may indicate the occurrence of non-tree-like patterns in the evolution of these taxa. Hybridization and introgression are phenomena shown to be related to the origin of some species and genera in bryophytes (see Natcheva and Cronberg 2004); thus, their possible role in shaping the patterns found in the Leucobryaceae cannot be disregarded.

Taxonomic treatment

Atractylocarpus subporodictyon (Broth.) Bonfim Santos & Stech, comb. nov. ≡ Dicranodontium subporodictyon Broth., Symb. Sin. 4: 20. 1929. ≡ Campylopus subporodictyon (Broth.) B.H.Allen & Ireland, Lindbergia 27: 76. 2002. ≡ Dicranum subporodictyon (Broth.) C.Gao & T.Cao, Bryobrothera 1: 218. 1992.—TYPE: China, “NW-Y.: An nassen Granitfelsen der wtp. St. im birm. Mons. bei Schutsche am Dijou-djiang (e Irrawadi-Oberlauf), 27°54', 2000 m. 7. VII. 1916” Handel-Mazzetti 9433 (holotype: H-BR; isotype: H).