We sampled 80 taxa of Orthotrichoideae, focusing on tribe Orthotricheae, with representatives of 7 of its 9 genera. Details on the samples included in the analyses are shown in Supplementary Table S1, with nomenclature following Tropicos (2022) database and abbreviations of authors of plant names following IPNI (2021) database. Specifically, we included 72 species of Orthotricheae, which constitutes approximately 30% of the accepted species: 26 species (out of 100 accepted) of Orthotrichum, 1 (of 2) of Nyholmiella Holmen & E.Warncke, 22 (of 70) of Lewinskya, 1 (of 4) of Pulvigera Plášek, Sawicki & Ochyra, 20 (of 70) of Ulota, 1 (of 1) of Plenogemma Plášek, Sawicki & Ochyra, and 1 (of 1) of Atlantichella.

To root the phylogenetic tree, we used representatives of tribe Zygodonteae as sister group, with 4 species of Zygodon Hook. & Taylor and 1 of Australoria F.Lara, Garilleti & Draper. As ultimate outgroup, we included 1 species of Leratia Broth. & Paris (of the sister subfamily Macromitrioideae).

We extracted DNA from the selected samples with a modified cetyltrimethylammonium bromide (CTAB) protocol (Doyle and Doyle, 1987) described in Breinholt et al. (2021). We used a Geno/Grinder 2010 mill (SPEX CertiPrep, Metuchen, New Jersey, United States) to lyse the cells and performed two rounds of chloroform washes followed by an isopropanol precipitation and an ethanol wash. We added 2 μl of 10 mg/ml RNase A (QIAGEN, Valencia, California, United States) to each sample between chloroform washes to remove RNA contamination.

We employed a target enrichment approach using the GoFlag 408 probe set to generate a multi-locus nuclear sequence dataset for phylogenetic analyses. The GoFlag 408 probe set targets 408 exons found in 229 single or low-copy nuclear genes and appears to recover many loci across mosses (Breinholt et al., 2021). Library preparation, target enrichment, and sequencing were done by RAPiD Genomics (Gainesville, FL, USA). Protocols for library preparation and hybridization are described in Breinholt et al. (2021). All enriched, pooled libraries were sequenced on an Illumina HiSeq 3,000 platform (Illumina; 2 × 100 bp).

We extracted the targeted loci from the raw sequence reads using a pipeline described in detail in Breinholt et al. (2021), and the scripts and reference sequences are available in Dryad (Breinholt et al., 2020). We trimmed the raw sequence reads with Trim Galore! Version 0.4.4¹ to remove adapters and bases with a Phred score below 20. We then assembled the targeted loci for each sample using iterative baited assembly (IBA; Breinholt et al., 2018), which conducts a de novo assembly with BRIDGER version 2014-12-01 (Chang et al., 2015) based on sequence homology of raw reads to a set of reference sequences for each locus. The IBA seeks to extend the assemblies beyond the target regions and recover as much of the more variable flanking intron regions as possible. In this sense, two types of matrices were generated as: (a) assembled sequences trimmed to the probe region (referred to as Probe Only matrices) and (b) full-length assembled sequences, i.e., including probe regions as well as flaking intron sequences (referred as Full Sequences matrices).

Next, we performed an orthology assessment using the target region sequences based on a tBLASTx (Camacho et al., 2009) search against nine flagellate plant genomes to remove potential paralogs. We also removed possible contaminants by performing a tBLASTx search of the assemblies for each locus against the reference sequences, and we removed any sequences that had the best hit that did not come from a moss. Finally, for each locus, we aligned the recovered sequences from the target regions only (Probe Only) and from the combined target and flanking regions from each locus (Full Sequences) with MAFFT 7.425 (Katoh and Standley, 2013). Putative isoforms from the same taxon were merged with a Perl script that used IUPAC ambiguity codes to represent putative heterozygous sites.

Although this pipeline does not explicitly phase loci, in some cases, a locus alignment might include multiple sequences from some samples, representing cases in which the BRIDGER assembler identified allelic diversity. In these cases, to reduce the possibility of including paralogs in our phylogenetic analyses, for each sample with more than one sequence, we removed all sequences from that sample from the locus alignment. Across the 405 loci from which we recovered sequences from at least four samples, we removed 1934 sequences, while retaining 25,544 single-copy sequences. Also, alignments of the flanking regions often have large gaps with sequences from one or a few samples due to indels and the high variability in the recovered length of the flanking sequence. Thus, to clean the alignment and reduce missing data, we ran a script to remove all columns from the locus alignments with fewer than four nucleotides. Additionally, the Probe Only matrix was also pruned with Gblocks (Castresana, 2000) with the following settings: -t = d -b1 = 51 -b2 = 60 -b3 = 8 -b4 = 8 -b5 = h. Summary statistics were calculated using AMAS (Borowiec, 2016). Finally, individual matrices from both datasets were concatenated into two independent supermatrices (i.e., Full Sequences supermatrix and Probe Only supermatrix).

Trees were inferred using two approaches: (a) a total evidence approach using maximum likelihood (ML) inference based on a concatenated matrix of all loci and (b) a summary species tree method that accounts for the Multiple Species Coalescent (MSC) with ASTRAL-III 5.7.8 (Zhang et al., 2018).

For the total evidence approach, we ran ML analyses of both the Full Sequences supermatrix and the Probe Only supermatrix. Phylogenetic analyses of these two datasets were executed in IQ-TREE 2.0.3 (Minh et al., 2020), after automatic model selection using ModelFinder (Kalyaanamoorthy et al., 2017) with the approximate likelihood ratio test (“-alrt” option). These analyses also included 1,000 bootstrap replicates and 1,000 ultrafast bootstrap (“bb” option). To investigate gene tree versus species tree concordance, we calculated two measures of genealogical concordance in our dataset, the gene concordance factor (gCF) and the site concordance factor (sCF), using the options “-gcf” and “-scf” in IQ-TREE. This approach provides a description of possible disagreement among loci and across sites within the sequence. We considered only branches with ultrafast-bootstrap support values >90% as statistically supported. Trees were plotted in FigTree 1.4.4.²

For the summary species tree approach under the MSC, individual gene trees were constructed using RAxML 8.2.12 (Stamatakis, 2014) applying GTR-CAT and 200 bootstrap replicates followed by slow ML optimization with the “-f a” option. Then, branches with BS < 50% were collapsed using Newick utilities (Junier and Zdobnov, 2010). Species tree inference under the MSC approach was then performed using ASTRAL-III, and branch support values were inferred through local posterior probabilities (LPP; Sayyari and Mirarab, 2016). Values of LPP > 0.95 were considered to represent strong branch support, although lower values (LPP = 0.7–0.9) also may indicate high support (Sayyari and Mirarab, 2016). To output quartet support values, we used the “-t 2” option. We plotted pie charts reporting the proportion of quartet values in R (R Core Team, 2020) using the packages ape (Paradis and Schliep, 2019), ggimage (Yu, 2021), ggtree (Yu et al., 2017), treeio (Yu et al., 2017), and their corresponding dependencies.

We categorized the studied species according to their niche preferences (regarding substrate and climate) and distribution. Orthotricheae mosses occur on three types of substrates: rocks, tree trunks (including large branches), and small branches (including twigs). All the species were characterized according to their preferences for one, two, or all three possible substrates, mainly on the basis of the expert knowledge of the authors and always recording the prevailing ecological behavior of the species, without considering the most exceptional situations (Mazimpaka and Lara, 1995). We classified the climatic preferences of each species regarding the degree of humidity of the climatic environment in which the species usually grow. These were also divided into three principal types: arid (with scarce precipitations and long periods of dry season, such as the Mediterranean climate), dry (with scarce to moderate precipitations, but without long periods of dry season), or wet (humid or hyper-humid climates, including local or regional situations with frequent mists that produce horizontal precipitations), and we characterized the species as showing preferences for one, two, or all three of them. Finally, we described the climatic preferences regarding the degree of thermicity according to the latitudinal bands where the species thrive as temperate (including cold-temperate), subtropical, and tropical-montane. Subtropical is used as defined by Troll and Paffen (1964) and includes warm climates, between tropical and temperate, with mean temperatures between 17 and 24°C, as prevail in the Mediterranean basin, southern California, The Cape Region, or Macaronesia. The characterization of the species was completed with a description of their geographical range: Subcosmopolite, Holarctic, Sub-antarctic, Australasia, Europe, Western Europe, Mediterranean basin, Macaronesia, East Africa, South Africa, East Asia, South India, North America, Central America, Caribbean, South America, Tropical Andes, and Patagonia.

The target enrichment recovered more than 200 nuclear loci (out of a possible 408) for 71 out of 80 samples, and more than 350 loci for 56 of these samples. Five samples largely failed, recovering 5 or fewer loci (Lewinskya acuminata, L. sordida, L. tasmanica, O. pilosissimum, and O. rivulare), and consequently, we did not include these samples in our analyses. Only five of the targeted loci were recovered in fewer than 10 samples. The average proportion of missing data in the Probe Only supermatrix was 0.63%, and only 10 out of 408 loci had more than 5% of missing data. The average proportion of parsimony informative sites per locus was 17.6%, and only 22 out of 408 loci had less than 10% of parsimony informative sites. The Full Sequences supermatrix was notably noisier, with an average proportion of missing data of 47%, and only 2 loci with less than 5% of missing data. Summary statistics are available in Supplementary Table S1 (sample statistics including number of loci and percentage of sequencing success) and Supplementary Table S2 (loci statistics including length, number of taxa, and number of variable, parsimony informative and missing data sites for both the Probe Only and the Full Sequences supermatrices). Raw data files are available in the GenBank Sequence Read Archive (SRA) under the BioProject number PRJNA819401. The unique accession number of each sample is available in Supplementary Table S3.

The analyses of the Full Sequences supermatrix and the Probe Only supermatrix yielded phylogenetic trees with similar topologies, although the Full Sequences trees showed shorter branches than the Probe Only trees for some of the nodes in the in-group. Therefore, the trees shown on Figure 1 and commented hereafter are those resulting from the Probe Only supermatrix analyses. The trees based on the Full Sequences supermatrix are included as Supplementary Figures, as well as pie charts reporting quartet support values for the Probe Only analyses.

Both species-based trees (IQ-TREE) and gene-based trees (ASTRAL) recover overall concordant clades (Figure 1). Many nodes are well supported with IQ-TREE ultrafastbootstrap (BS; i.e., values greater than 90%) and ASTRAL posterior probability (LPP; i.e., values greater than 0.9). Some of the branches that receive low BS or LPP value support have low gCF scores and/or have low quartet scores (Supplementary Table S4).

Samples of Orthotricheae were included in a single clade sister to samples of Zygdonteae in all the analyses. Within the Orthotricheae clade, samples were distributed in two monophyletic groups, one including the samples of Orthotrichum and Nyholmiella (not well supported clade E, dark green colored in Figure 1) and the other including the samples from the remaining genera (highly supported clade A, light green). Within clade E, all the analyses placed the sample of Nyholmiella as sister to a highly supported clade F, which included all the samples of the species of Orthotrichum. This Orthotrichum clade F was in turn divided into subclades according to both the species- and gene-based trees, although not all these inner clades were maximally supported, and there was some incongruence on the grouping of Orthotrichum patens, O. stramineum, O. alpestre, O. consobrinum, O. cucullatum, and O. consimile.

Regarding clade A, all the analyses established an inner subgrouping in three large and strongly supported clades, named B, C, and D in Figure 1. The relationships of these three clades varied depending on the analysis: the species-based tree supported a closer relationship between clades C and D, and placed B as sister of the two, whereas the gene-based tree supported a sister relationship for B and C, and placed D as sister of them. Clade B included all the samples of Lewinskya and was in turn divided into inner subclades both according to the species- and gene-based trees, although this inner grouping was not maximally supported in any of the analyses. The groups resulting from both the species- and gene-based trees were overall congruent, except for the grouping of Lewinskya speciosa, L. affinis, L. iberica, and L. firma. Similarly, clade D was divided into subclades (not fully supported) that were overall congruent among the species- and gene-based trees, except for the relationships established for Ulota longifolia, U. japonica, U. bruchii and U. hutchinsiae. This clade D included all the samples of Ulota with the exception of the sample of the species U. bellii, which was included in clade C together with the small genera Plenogemma, Pulvigera, and Atlantichella.

The ecological and biogeographical affinities of the studied species were plotted in the phylogenetic framework to assess whether the recovered clades could reflect ecological or biogeographical patterns. As explained above, the main clades (i.e., clades representing the genus taxonomic level) established by all the analyses performed were congruent, even though the relationships among them sometimes differed. This was the case for the position of clade C, which includes three taxa (Plenogemma phyllantha, Atlantichella calvescens, and Ulota bellii) that have been traditionally treated as Ulota due to their morphological similarities. These similitudes were supported by the IQ-TREE analysis, which showed a sister relationship of clade C and Ulota. For this reason, we selected the IQ-TREE phylogenetic reconstruction to plot the ecological and biogeographical affinities, in order to identify the possible patterns underlying the recovered clades (Figure 2).

Most of the established clades were clearly congruent with the substrate and climatic preferences. Thus, Lewinskya species as a whole tend to be specialized in colonizing tree trunks and large branches, although most of them often also grow on small branches and twigs. Regarding humidity, species of this genus tend to show preferences for dry climate, although some species are typical or common in wet areas, and L. rupestris, a cortico-saxicolous species, shows a wide range of humidity tolerance. The highest variability is found among the temperature preferences that range from cold-temperate to tropical-montane at the genus level. Notably, this variability agreed in general terms with the inner subclades of Lewinskya, even though these subclades were not always maximally supported. For example, the maximally supported clade including L. tasmanica and related species could be defined as typically temperate, whereas the not maximally supported clade of L. arborescens and related species could be defined as typically tropical-montane. Conversely, the clades recovered did not show a clear geographical pattern, and species that usually coexist, such as L. breviseta and L. iberica in the Mediterranean or L. arborescens and L. firma in East Africa, were not closely related.

The clade including the species of Ulota was also clearly differentiated from the rest by its ecological preferences (Figure 2). Regarding substrate affinities, this group of species shows a clear tendency toward small branches and twigs (except for the mainly saxicolous U. hutchinsiae), although most of them can also appear on large branches and trunks. As for climatic preferences, these species are all typical of wet areas, temperate, or cold-temperate, from both hemispheres. Exceptionally, they can also thrive in subtropical (Lara et al., 2022) or tropical-montane areas (Garilleti et al., 2015). Again, no clear biogeographical pattern could be established for this genus.

The third large genus of the family is Orthotrichum. As shown in Figure 2, this genus is more variable regarding the ecological preferences of its species than the two above mentioned, although some ecological patterns were also obtained, especially for the inner subclades. Thus, this genus includes species usually growing on rocks (some of which were grouped in the subclade including O. anomalum and related species), on trunks and large branches (such as those included in the subclade around O. subexsertum), and (more rarely) on small branches and twigs. Regarding its climatic preferences, it ranges from an affinity for wet areas to dry or even arid ones and from cold-temperate to tropical-montane areas. No clear pattern was recovered that could easily explain the phylogenetic clades recovered, either based on the climatic or biogeographical affinities.

The remaining five genera of Orthotricheae included in this analysis were represented by a single species each. Four of these were grouped into clade C that, according to the species-based reconstruction, is sister to Ulota (Figures 1, 2). This clade C did not represent any biogeographical pattern, since, e.g., Atlantichella calvescens and Ulota bellii, inferred to be sister taxa in all the analyses, show an antipodal distribution. Conversely, they share a clear preference for small branches and twigs, and climatic preferences for wet areas, from cold-temperate extending to the subtropical latitudes in the case of A. calvescens.

The different analyses performed (Probe Only supermatrix vs Full Sequences supermatrix; species-based trees vs gene-based trees) yielded overall congruent results and well resolved phylogenetic reconstructions. This adds evidence to the utility of the GoFlag enrichment probe set (Breinholt et al., 2021) to resolve not only phylogenetic relationships across distantly related taxa, but also among more closely related taxa, as it has been demonstrated for other groups of flagellate land plants such as ferns (e.g., Fawcett et al., 2021).

Nevertheless, there was discordance regarding the position of some species depending on the analyses (i.e., topology of the ML supermatrix trees vs ASTRAL species trees). There are numerous reasons that can explain this type of conflict, including: (i) short branch lengths (i.e., not enough molecular evidence to be able to resolve the phylogenetic relationships among taxa), (ii) incomplete lineage sorting or another type of genuine conflicting genealogical histories (see a revision in, e.g., Degnan and Rosenberg, 2009), or (iii) a selection of loci that are not informative or that provide unclear information (phylogenetic noise, sensu Straub et al., 2014). In our case, most of the nodes that produce conflict are associated with short branches, which is especially evident in the case of the inner nodes of Lewinskya (clade B in Figure 1) and the position of clade C in relation to clades B and D. The branch suspending the relationship between clade C and clades B-D is very short (due to lack of information, a quick event of diversification, etc.) and thus, some uncertainty involves it: it might appear as sister to either clade B or clade D.

In addition, some of the branches that receive low BS or LPP support have as well low quartet scores, indicating gene tree conflict, and/or have low gCF scores, which reflects that few gene trees support the grouping. For instance, clade E has low BS and LPP support. For this node, which corresponds to a very short branch in both trees, the three quartets have similar values (q1 = 0.39, q2 = 0.29 and q3 = 0.31) and the gCF/sCF are low (14 and 33%, respectively).

In this scenario, it is necessary to further analyze the poorly supported relationships before making conclusions that involve the conflicting nodes within Lewinskya, Ulota, and Orthotrichum. Future studies will include designing a specific target enrichment set for this family, since this could potentially decrease phylogenetic noise and missing data, as it has been previously demonstrated for other groups such as the tribe Cardueae of Compositae (Herrando-Moraira et al., 2019) or the Cyperaceae (Larridon et al., 2020).

Our phylogenetic reconstruction is the most complete so far published for the tribe Orthotricheae, both in terms of number of taxa included and number of loci analyzed. The results obtained are overall congruent with those published by Draper et al. (2021) and Wang et al. (2021). As proposed by Draper et al. (2021), we confirm that Orthotricheae contains two subtribes, Orthotrichinae F.Lara, Garilleti & Draper and Lewinskyinae, which correspond to clades A (maximally supported) and E (not maximally supported) in Figure 1. Moreover, Draper et al. (2021) proposed the segregation at the genus level of both Australoria (separate from Zygodon, in Zygodonteae) and Atlantichella (separate from Ulota), which is also supported by our results (Figure 1).

According to our phylogenetic reconstructions, the genus Ulota as currently conceived remains polyphyletic, since U. bellii is placed in a separate fully supported clade (namely C, Figure 1) from the rest of the Ulota species included in this study (grouped together in a monophyletic and maximally supported clade D in Figure 1). Ulota bellii shows a characteristic combination of morphological characters that also justifies its segregation from Ulota in a separate genus that we propose to name Rehubryum F.Lara, Garilleti & Draper. A brief discussion of the diagnostic morphological characters is provided in the taxonomical description section.

None of the phylogenetic reconstructions so far published has been able to fully resolve the relationships of the different genera within the two subtribes, since many of the clades lacked support. In addition, the relationships suggested by previous studies pointed to incongruent results. Based on their 6-loci results, Wang et al. (2021) suggested that, within Lewinskyinae, Ulota is sister to Lewinskya and that these are grouped with Plenogemma and Pulvigera in an unresolved polytomy. Noteworthy, these authors did not include the genera Atlantichella and Rehubryum in their study. Conversely, Draper et al. (2021) considered Plenogemma as sister to Ulota, and both of them were grouped in a polytomy with Lewinskya and Atlantichella, based on a selection of 4 different loci and without representation of Rehubryum. Regarding Lewinskyinae, we obtained different topologies depending on the analyses, but all our results point to a sister relationship of Plenogemma and Pulvigera, as well as of Atlantichella and Rehubryum, and these four genera are assembled in a monophyletic maximally supported clade C (Figure 1). The sister relationship of this clade, with either Ulota (species-based trees) or Lewinskya (gene-based trees), remains ambiguous. Our results fail to provide final evidence regarding the phylogenetic relationships for the genera within Orthotrichinae, since we lack data for Stoneobryum D.H.Norris & H.Rob. and Sehnemobryum Lewinsky & Hedenäs, so further studies are needed to reach final conclusions. Nevertheless, our results point to a different solution than those suggested in the previously published phylogenies and stress the need to further explore this group of taxa to unravel the intergeneric relationships. On one hand, it is necessary to include all the genera of Orthotricheae in a complete phylogeny to resolve the relationships within Orthotrichinae. On the other hand, there is a need to obtain additional molecular data that could help to discern the evolutionary history of the group. As an example, the conflicting solutions suggested by this study (based on nuclear loci) and those previously published (which include data from organellar genomes) could reflect a complex evolutionary history with ancient hybridization events, as it has been observed in other groups such as algae (e.g., Bringloe et al., 2021), angiosperms (e.g., Bogdanova et al., 2021), or other bryophytes (Meleshko et al., 2021).

In any case, our results indicate a puzzling biogeographic history for the extant taxa, given the strongly supported close relationship of Plenogemma, Pulvigera, Atlantichella, and Rehubryum shown by all the analyses (clade C, Figure 1). These four taxa include hyperoceanic mosses, but they highly differ in their distributions, reproductive strategies, and morphology (Figure 2). Pulvigera comprises four species with Orthotrichum-like aspect, all of them found in westernmost North America, although one species can also be found in some Pacific archipelagos, and another one is present in western Europe and the Mediterranean (Lara et al., 2020). All species of Pulvigera are dioicous mosses with no specialized vegetative reproduction, except for P. lyellii, the one with a disjoint Holarctic distribution, which generates abundant gemmae for vegetative propagation. Plenogemma phyllantha, the only representative of its genus, is an Ulota-like moss with dioicous distribution of sexes that reproduces mainly by vegetative propagules. It shows a wide and irregular bipolar distribution, involving most continents and several oceanic archipelagoes, including some remote sub-Antarctic islands (Garilleti et al., 2015). In turn, both Atlantichella and Rehubryum are monotypic genera comprising Ulota-like, monoicous mosses that actively reproduce sexually and lack any type of specialized propagules for vegetative reproduction. Atlantichella calvescens is an endemic of the northeastern Atlantic area, found in the Macaronesian archipelagoes, British Isles, and scattered localities on the western coast of Europe and the Mediterranean basin (Lara et al., 2022), whereas Rehubryum bellii is only known from the Antipodes, restricted to New Zealand.

The close relationship of the four taxa comprising clade C agrees with some morphological similarities (see below) but raises a question about the aspect, sexual system, and distribution of their common ancestors and the evolutionary history of the group. The appearance of the ancestral taxon could be either of the two shown by the current descendants, since an original Ulota-like appearance would agree with the topology established by the species tree phylogeny, while the Orthotrichum-like appearance would be supported by the topology of the gene-based tree. The two possible sister taxa (Lewinskya and Ulota s.s.) are entirely monoicous lineages, so the dioicous condition of the subclade formed by Plenogemma and Pulvigera would in any case be a derived feature, whereas the monoicous condition of the subclade formed by Atlantichella and Rehubryum would coincide with that of the hypothetic ancestor. We can hypothesize that both the original ancestor and those in the origin of the two main subclades must have been species with high dispersal capacities, as presently shown by many Orthotrichoideae (Vigalondo et al., 2016, 2019). Thanks to recurrent dispersal events, they must have been able to colonize distant hyperoceanic areas of the planet.

In addition to the relationships among the genera of Orthotricheae, this study provides data regarding the infrageneric grouping within the most speciose genera of the tribe. Several infrageneric proposals have been made based on morphological resemblances, all of them focusing exclusively on Orthotrichum sensu lato (for a summary, see Lewinsky-Haapasaari and Hedenäs, 1998). Our results suggest that Lewinskya, Ulota, and Orthotrichum can be subdivided into several groups: at least three clades could be recognized within Lewinskya (although this grouping is not maximally supported and depends on the analysis performed, species- or gene-based trees); also samples of Ulota sensu stricto are distributed in at least three clades, although two of them are not maximally supported; and four groups are established within Orthotrichum, all of them maximally supported although their sister relationships partly vary depending on the analyses. Noteworthy, none of these clades is fully congruent with those currently in use (Vitt, 1973; Lewinsky, 1993; Lewinsky-Haapasaari and Hedenas, 1998), although many of the clades here suggested reflect either morphological similarities or ecological preferences (Figure 2). As an example, one of these clades unites the most xerophytic taxa included in the analysis (namely, O. macrocephalum, O. diaphanum, O. pallens, O. subexsertum, O. rogeri, O. scanicum, O. bartramii, O. tenellum, and O. shevockii), while another includes taxa that share stomata located in the lower part of the capsule and a hairy vaginula (O. patens, O. stramineum, O. alpestre, and O. consobrinum). Similar results pointing that the traditionally accepted subgenera do not reflect natural phylogenetic groups have been obtained in previous studies (e.g., Goffinet et al., 2004; Sawicki et al., 2012). Unfortunately, this study lacks a complete representation of the diversity of Orthotrichum (represented here by 26 of 100 species), Lewinskya (22/70), and Ulota (20/70), and so the present results are too preliminary as to already propose any new infrageneric division. More data are also needed to increase the resolution of the groups and to be able to infer their taxonomic status.

There is a generally accepted assumption that the epiphytic environment constitutes a hostile one for the development of plants, mainly due to drought stress and restricted nutrient supply (e.g., Pugnaire and Valladares, 1999). Nevertheless, it has been also argued that the epiphytic environment can as well be considered as an available space with unexploited resources (Lüttge, 2008) and with a high diversity of microhabitats due to different gradients of light, temperature, humidity, nutrient supply, and substrate characteristics related to bark structure and branch demography (e.g., Zotz, 2016). This has been especially analyzed in tropical forests (e.g., Lüttge, 2008) and on epiphytic vascular plants (Zotz, 2016). A revision synthesizing the underlying biotic interactions that can have been important for epiphyte ecology and evolution has been recently published (Spicer and Woods, 2022). In this study, the authors highlight the importance of acquiring unique adaptive traits to thrive in fine-scale microhabitats within the epiphytic environment, as evolutive drivers in some vascular epiphyte groups. This has been especially claimed for orchids (e.g., Givnish et al., 2015) and bromeliads (e.g., Benzing, 1987), but little has been published on non-vascular plants (Spicer and Woods, 2022) and the specific factors that promote diversification in mosses are not yet well known.

Bryophytes are poikilohydric organisms whose behavior and adaptations to drought stress strongly differ from those of vascular plants (Barkman, 1958). Mosses and other bryophytes compensate the absence of an impermeabilizing epidermis with the ability to enter in a dormancy state that enables them to tolerate desiccation for long periods, whereas water uptake is mostly ectohydric. According to Huttunen et al. (2018), morphological characteristics connected to ectohydry may be driven by adaptations to environmental conditions. This could be interpreted as evidence of how acquiring adaptive traits can drive diversification in bryophytes, which has been suggested for orchids and bromeliads.

Orthotrichaceae is one of the most speciose bryophyte families and has diversified mostly in the epiphytic environment. Within it, the very species-rich subfamily Macromitrioideae has diversified in warm tropical epiphytic environments where also most vascular epiphytes grow. But to what extent the diversification of the tribe Orthotricheae, which specialized in temperate environments (including high tropical altitudes), can be interpreted in the same terms of adaptation to a wide variety of meso- and microenvironmental conditions is something that has not been previously addressed. Previous works lacked support to approach this question, but our results have revealed a clear ecological pattern involving substrate and climatic preferences in the major clades recovered that make progress toward identifying possible drivers of diversification. Conversely, we have not been able to detect a clear biogeographical signal in the phylogenetic reconstructions. This could indicate the evolutionary importance of acquiring adaptative traits that enable the colonization of certain epiphytic microhabitats. In this context, the diversification at the genus level in the tribe Orthotricheae could be partially explained by the adaptation to a certain combination of the substrate characteristics (rocks, trunks, or small branches) together with climatic preferences regarding humidity and temperature. This could also explain an infrageneric diversification, although further data and the inclusion of a wider representation of species are needed to be able to safely achieve conclusions at this taxonomical level. Finally, as suggested by Huttunen et al. (2018) and in line with the results by Draper et al. (2021) on the prevalence of homoplasy among Orthotricheae, the importance of the adaptation to the environment could explain the parallel morphological evolution of the different genera specialized on the epiphytic habitat. Moreover, the results of our study point toward the idea that, at least for bryophytes, stressful environments can promote diversification and harbor great diversity.

The genus Rehubryum is proposed to accommodate Ulota bellii Malta on the basis of its peculiar combination of morphological traits and phylogenetic position.

Type: Rehubryum bellii (Malta) F.Lara, Garilleti & Draper, comb. nov. ≡ Ulota bellii Malta, Acta Horti Bot. Univ. Latv. 7: 15. 1933.

Diagnosis: Plants autoicous, forming cushions. Leaves spirally arranged, strongly crisped when dry, lanceolate, gradually dilated to a base scarcely concave, often plicate on both sides of the nerve, with margins plane or erect-incurved in one side, especially in the transition between base and lamina, leaf lamina unistratose and mainly plane at margins; basal cells long rectangular to linear, somewhat sinuous, with thickened walls; basal-marginal cells differentiated, hyaline, quadrate to rectangular, with thickened transverse walls, forming a narrow marginal band along the base and proximal end of the lamina; margins at upper base with papillose teeth arising at the junctions between two cells; submarginal rows of elongated cells differentiated from base through lower third of the lamina; median and upper leaf-cells rounded to elliptical, with low papillae. Propagula absent. Perichaetial leaves slightly differentiated, with a broader base. Seta 3–5 mm long, twisted counterclockwise. Capsule exserted, oblong-ovoid to short cylindrical, symmetric, entirely ribbed. Exothecial bands narrow differentiated from mouth to urn base. Stomata superficial, at urn base and neck. Peristome double; exostome of 8 pairs of teeth, easily splitting after recurving; endostome of 16 linear segments, involute when dry, with a low connective membrane. Operculum rostrate, with base almost plane. Spores unicellular, isomorphic, papillose. Calyptra mitrate, with abundant stout hairs.

Etymology: rehu is a Maori noun for mist but also a verb that means to pass out of sight, disappear, and render unconscious (Moorfield, 2022), all of which seems appropriate for this moss that lives in foggy environments and has gone virtually unnoticed as a different genus.

The New Zealand endemic Rehubryum bellii appears to be a typical species of the genus Ulota as it shows the general look that most of these mosses have (Figures 3A,B), as well as many of the details that serve as morphological characters for their taxonomic characterization. Indeed, in his recent review of Ulota in New Zealand, Fife (2017) considers U. bellii not worthy of taxonomic recognition and synonymized it with U. lutea (Hook. f. & Wilson) Mitt. However, R. bellii is easily separated from any species of Ulota in the Australasian area by the possession of an endostome consisting of 16 filiform segments, involute when dry, all of them well developed (Figure 3C). Other differential characters, such as the oblong-ovoid shape of the capsule (Figure 3B) or the possession of an exostome with 8 pairs of teeth easily splitting (Figure 3C), have already been highlighted since the description of the species (Malta, 1933; Sainsbury, 1955). However, this moss has two additional very significant characters at the distal part of the leaf base (Figure 3D): (a) submarginal bands of elongate cells ascending from the transition base-blade some way up and (b) margins of some leaves denticulated by prominent papillae arising at the junction between every two marginal cells. Both characters seem to have gone unnoticed and their discovery while examining our New Zealand collections was fundamental for the inclusion of samples of this species in the phylogenetic analysis. In fact, in a previous study (Draper et al., 2021), the occurrence of leaves with submarginal bands of elongate cells was revealed as a characteristic shared by the genera Atlantichella and Plenogemma, whereas basal leaf margins with geminate teeth are also found in these two genera and in Pulvigera, where the feature is especially visible. In contrast, both traits appear to be absent in Ulota. The phylogenetic reconstructions obtained in the present study group in the same clade all the four genera in which these features have been so far observed, giving these characteristics an unsuspected taxonomic significance.

The distinction between Rehubryum bellii and Atlantichella calvescens does not entail any difficulty because there are many morphological differences between the two species, especially in the sporophyte. Thus, for example, whereas A. calvescens has capsules broadly ribbed and strongly contracted below mouth when dry, with an endostome of 8 linear segments and devoid of connective membrane, R. bellii has capsules finely ribbed, not contracted below mouth when dry, with an endostome of 16 filiform segments and with connective membrane. As these are the only known species of these two genera, it could be thought that their differential traits also serve to characterize Rehubryum versus Atlantichella. However, all the morphological characters that serve for differentiating both species vary within the large genus Ulota (Caparrós et al., 2014; Caparrós, 2015), so their value for characterizing genera among the Lewinskyinae is doubtful. It should also be noted that Draper et al. (2021) demonstrated that within this group of mosses most characters used for separating genera are homoplastic.