Grammitid ferns comprise a well-supported clade of ~900 primarily epiphytic, tropical species of the widespread family Polypodiaceae [1] (Parris, unpublished data). One of the long-standing questions in the study of grammitid ferns has been the circumscription of genera, with some authors recognizing a single genus [2] and others up to 24 [3] . The first broad molecular phylogenetic study of grammitids provided new insights and some clarification of generic delimitations. Ranker et al. [4] found that only six of 12 genera for which more than one species was sampled were recovered as monophyletic. Large genera in both the Paleotropics (e.g., Ctenopteris Blume ex Kunze s.l., Grammitis Sw. s.l.) and the Neotropics (e. g., Lellingeria A. R. Sm. & R. C. Moran, Terpsichore A. R. Sm.) were found to be polyphyletic. Ranker et al. [4] cited convergent evolution and the use of homoplastic characters by taxonomists to account for the problems in generic circumscription. That study has been followed by more intensive studies of particular grammitid groups (e.g., [5] [6] [7]), some of which have led to nomenclatural changes, including the recognition of new genera [8] [9] [10] [11] [12] [13] [14].

We focus here on the Lellingeria myosuroides (Sw.) A.R. Sm. & R.C. Moran group recognized by Smith et al. [15]. Phylogenetic studies of molecular and morphological data provide strong support for the monophyly of this group [16] [5] [4] [7] and its sister relationship to a clade containing all remaining species of Lellingeria (the Lellingeria s.s. clade of Labiak et al., [5]). The group includes 24 species, which are distinguished from Lellingeria s.s by segments with a single unbranched vein, the fertile veins with dark sclerenchyma visible beneath the sporangia, linear hydathodes, and a single sorus per segment (Fig. 1) and will be recognized as a new genus by Labiak [10]. Unlike Lellingeria s.s. , which is restricted to the Neotropics, the L. myosuroides clade has a pantropical distribution, and is found in the Neotropics (West Indies, Mexico to Bolivia, and Brazil), Hawaiian Islands, French Polynesia, Africa, Madagascar, and La Réunion Island [5]. Estimations of relationships within the L. myosuroides clade, including attempts to explain biogeographical patterns, however, have varied across studies. For example, the focus of the work of Geiger et al. [16] was to use molecular phylogenetics to explore the geographical origins of multiple Hawaiian fern lineages, including the Hawaiian endemic L. saffordii (Maxon) A. R. Sm. & R. C. Moran (Fig. 1e, 1f). Their results supported a sister relationship between L. saffordii and the French Polynesian endemic L. subcoriacea (Copel.) A.R. Sm. & R.C. Moran, with that clade being sister to the Neotropical L. limula (Christ) A.R. Sm. & R.C. Moran. The clade of those three species was embedded in a larger clade of Neotropical species. Thus, Geiger et al. [16] inferred a single dispersal event from the Neotropics of the common ancestor of the two Pacific species. By contrast, Labiak et al. [5] and Sundue et al. [7] included additional species of Lellingeria in their analyses, although the latter did not include L. subcoriacea . Both of those studies provided robust support for a sister relationship between the Hawaiian L. saffordii and the Mexican L. hellwigii (Mickel & Beitel) A.R. Sm. & R.C. Moran. Although the analysis of Labiak et al. [5] could not resolve the exact relationships of L. subcoriacea , it was clearly not sister to L. saffordii or even the L. saffordii-L. hellwigii clade. Thus, they hypothesized two independent dispersal events of the ancestors of L. saffordii and L. subcoriacea from the Neotropics.

We present new evidence of relationships and biogeographical affinities among members of the L. myosuroides clade based on phylogenetic analyses of plastid DNA sequences.

Examples of species in the Lellingeria myosuroides clade.

A. Habit of L. pumila (Brazil). B. Segments and sori of L. limula, showing the green spores inside the sporangia (Costa Rica). C. Habit of L. wittigiana (Brazil). D. Fertile portion of the lamina in L. wittigiana (Brazil). E. Habit of L. saffordii (Hawaii). F. Fertile and sterile portion of the lamina in L. saffordii (Hawaii). (A, C, and D by P. Labiak; B by R. C. Moran; E and F by T. A. Ranker).

Taxonomic sampling — Outgroup sampling included 18 species (19 accessions) including species of Lellingeria s.s., Melpomene A. R. Sm. & R. C. Moran , the Terpsichoreanfractuosa clade, and the Terpsichoresubscabra clade. These were chosen because they have been recovered as the closest relatives to the Lellingeriamyosuroides clade in previously published studies [4] [5] [7]. The ingroup is represented by 13 species (24 accessions), representing 54% of the known species. Voucher information is listed in Appendix 1.

DNA extraction, amplifications and sequencing — Genomic DNA was isolated following the protocol detailed in Ranker et al. [17]. The three plastid DNA markers used for the analyses ( atpB , rbcL , and trnL-trnF ) were amplified by PCR following the protocols detailed in Ranker et al. [17]. All 28 newly obtained consensus sequences have been submitted to GenBank (Appendix 1).

Alignment and phylogenetic analyses — Consensus sequences were automatically aligned using the program Muscle v3.6 [18], and the resulting alignment was manually checked and revised when necessary. Following alignment, we coded the resulting gaps for the trnL-trnF region following the simple coding model as suggested by Simmons and Ochoterena [19], using the program 2xread [20]. Data matrices were constructed using the program ASADO [21] and analyzed using equally weighted maximum parsimony (MP), and Bayesian inference (BI). Maximum parsimony analyses were performed using the program TNT v1.1 [22]. For all MP analyses, heuristic searches were performed with 10,000 parsimony ratchet replicates ([23]; 200 iteration ratchet, the up and down weights set to 5% each), holding 20 trees per ratchet, followed by tree-bisection-reconnection (TBR)-max branch swapping. Support for nodes for the combined data set analysis was calculated by bootstrap analyses (BS), with 1000 replicates doing 10 ratchets per replicate, holding 20 trees per ratchet, and keeping only the strict consensus. The best performing evolutionary model was obtained by the Akaike information criterion (AIC; [24]), using the program jModeltest v.0.1.1 [25]. Bayesian analyses (BI) were performed using the program MrBayes v.3.1.2 [26] [27] on the freely available Bioportal (www.bioportal.uio.no). The coded gap characters (for trnL-trnF ) were included and analyzed separately from the rest of the sequence data, being set to follow the model implemented in MrBayes for binary data: “lset coding = variable”. The partition strategy included the models for each marker, separate codon positions for atpB and rbcL , and the binary model for the coded gap characters. Three independent runs were started from random trees, consisting of four chains each, one cold and three hot, with the temperature parameter set to 0.2. The analyses were run for 10 million generations, sampling every 1000th generations. To assess whether the Markov chain Monte Carlo (MCMC) reached stationarity after the “burn-in” period, we compared the standard deviation of split frequencies, and by looking at the posterior probability of each clade, as suggested by Huelsenbeck et al. [28] , using the online program AWTY [29]. Our data file has been submitted to TreeBase under study number S11022 (http://purl.org/phylo/treebase/phylows/study/TB2:S11022) .

Nomenclatural novelties ― The new species mentioned in the text are not intended to be formal proposals. They will be validly published in subsequent articles.

We report here only the results of the analyses of the combined dataset, the resulting phylogenetic trees of which were more resolved and statistically more strongly supported than each tree resulting from the three separate data matrices. The MP analysis recovered three equally parsimonious trees, with L = 628, CI = 0.64, and RI = 0.87. These collapsed at three nodes in a Nelson strict consensus tree. Bootstrap support values (BS) are shown on the strict consensus tree (Fig. 2). Convergence between the three runs of the Bayesian analysis was satisfactorily achieved

Strict consensus tree from parsimony analysis of the combined dataset.

Numbers above branches are bootstrap values. Names in black text are of Neotropical species. Other text colors represent the geographical areas indicated.

The parsimony analysis also found strong support (100 BS) for a sister-taxon relationship between the L. saffordii-L. hellwigii clade and what had been called the L. myosuroides clade, as had Labiak et al. [5], but with some significant differences because of the addition of two new accessions. Labiak et al. [5] included data from six South American/Andean accessions labeled L. myosuroides, however, recent observations [10] have shown that these Andean specimens should be referred to Xiphopteris jamesonii Hook., a species that has been considered as a synonym of L. myosuroides by recent authors (e.g., [15] [30]. This species occurs in Central America and the Andes and is morphologically and geographically distinct from L. myosuroides , which is restricted to the Antilles. Our dataset included all of the sequence data used by Labiak et al. [5], plus a new accession of true L. myosuroides from the Dominican Republic. We also included a new accession from La Réunion Island in the Indian Ocean, which had originally been identified as L. myosuroides but will be described soon as a new species, and that is restricted to La Réunion (Parris, unpublished data). We tentatively refer to it as L. reunionensis Parris sp. nov. ined. Table 1 presents some of the key differences among X. jamesonii, L. myosuroides, and L. reunionensis. Our parsimony and Bayesian analyses placed both L. myosuroides and L. reunionensis within the X. jamesonii clade, rendering X. jamesonii paraphyletic. Lellingeria myosuroides was supported as sister to an accession of X. jamesonii (98 BS, 1.00 PP) and L. reunionensis was placed with 100 BS and 1.00 PP support as a member of a distinct clade with four other accessions of X. jamesonii. One interpretation of this is that X. jamesonii is the ancestral taxon for L. myosuroides and L. reunionensis, both of which diverged from X. jamesonii subsequent to independent long-distance dispersals away from the Andes. Alternatively, there could be multiple cryptic species within this clade for which distinctive characters have yet to be discovered.

Another interesting pattern is the fact that the L. myosuroides clade is geographically highly diverse, having both Neotropical and Paleotropical species, including some extremely localized endemics, whereas its sister clade (Lellingeria s.s.) is more species rich but restricted entirely to the Neotropics. We suggest that detailed studies of life history and other attributes may be of value in helping to explain this phenomenon.

Table 1. Key characters distinguishing X. jamesonii, L. myosuroides, and L. reunionensis.

The authors have declared that no competing interests exist.