Toward a monophyletic Cheilanthes: The resurrection and recircumscription of Myriopteris (Pteridaceae)

Abstract The fern genus Cheilanthes (Pteridaceae) has perplexed taxonomists for more than two centuries. Complex patterns of evolution involving rampant morphological convergence, polyploidy, hybridization, and apomixis have made the taxonomy of this group especially difficult. Fortunately, recent phylogenetic analyses have helped to clarify relationships among cheilanthoid taxa. Based on these findings, we here formalize an updated taxonomy for one monophyletic clade comprising 47 primarily North and Central American taxa usually included in Cheilanthes. Because the type species of Cheilanthes (Cheilanthes micropteris) is only distantly related to this clade, we resurrect the genus Myriopteris to accommodate these taxa, and present a revised circumscription for the group, including 36 new combinations.


Introduction
A "practical and natural" generic classifi cation of cheilanthoid ferns (Pteridaceae) has eluded taxonomists for more than 200 years and was viewed by Tryon and Tryon (1982) as one of the most contentious issues in fern systematics. Central to the problem is the circumscription of the large genus Cheilanthes , which all molecular studies with suffi cient sampling indicate is polyphyletic (see Gastony and Rollo 1998;Kirkpatrick 2007;Prado et al. 2007;Schuettpelz et al. 2007;Zhang et al. 2007;Rothfels et al. 2008;Eiserhardt et al. 2011). Since the initial description of Cheilanthes (Swartz 1806) encompassing 16 species, various authors have moved hundreds of taxa into (e.g., Domin 1913;Mickel 1979) and out of (e.g., Fée 1852;Smith 1875;Ching 1941) the genus. Of the ca. 500 validly published species names in Cheilanthes , some 60% have, at some point, resided in other genera. Th e lack of defi nitive taxonomic characters in this group often is attributed to widespread convergent evolution in the drought-prone habitats occupied by these ferns Tryon 1973, 1982), and the problem is likely insoluble based on morphology alone. However, the same genetic evidence that highlights shortfalls in the current classifi cation provides a key to solving this puzzle. As DNA sequence data proliferate and morphological features are reexamined in light of molecular phylogenies, it eventually becomes possible to recognize monophyletic assemblages of species that can be circumscribed as genera. We now have reached this point with certain groups of cheilanthoid ferns, at least in terms of removing taxa and clades that cannot reasonably be included within Cheilanthes (Link-Perez et al. 2011;Li et al. 2012).
Here, we focus on the primarily New World lineage previously referred to as the "American Cheilanthes " (Kirkpatrick 2007), myriopteroid (Rothfels et al. 2008), or myriopterid (Windham et al. 2009;Eiserhardt et al. 2011) ferns. Limited sampling in each of those analyses indicated that these ferns might represent a well-supported, monophyletic group, an assumption fully supported by the more complete (85%) taxon sampling of Grusz et al. (in review). In addition to suggesting the monophyly of the myriopterid lineage, the analyses of Rothfels et al. (2008) and Eiserhardt et al. (2011) conclusively demonstrated that this clade was quite distantly related to the type species of Cheilanthes , C. micropteris (results summarized in Fig. 1). Th is improved understanding of phylogenetic relationships among cheilanthoid ferns necessitates a taxonomic revision that can be achieved by one of two options: 1) all taxa derived from the most recent common ancestor of C. micropteris and the myriopterid ferns could be assigned to a single genus (which would not be called Cheilanthes because of the priority of Hemionitis ), or 2) myriopterid ferns could be transferred to a diff erent genus, refl ecting their phylogenetic distinction from Cheilanthes s.s . Th e fi rst option would require 400+ new combinations in Hemionitis (or the conservation of Cheilanthes against it followed by more than 100 new combinations in that genus). It would also subsume a number of cohesive, well-characterized genera that are clearly distinct based on morphological, molecular, and cytological grounds, including Adiantopsis (Link-Pérez et al. 2011), Argyrochosma (Windham 1987;Sigel et al. 2011), Astrolepis (Beck et al. 2010), Doryopteris (Yesilyurt 2004), Gaga (Li et al. 2012), and Notholaena (Rothfels et al. 2008). Th is approach would maximize the number of nomenclatural changes while simultaneously obscuring well-documented phylogenetic relationships, resulting in the inclusion of all but six cheilanthoid species in one genus. Because we consider this option untenable, we have, instead, chosen to remove the myriopterid ferns from Cheilanthes .
When any species or clade is removed from Cheilanthes , the fi rst issue that must be addressed involves their relationship to Allosorus pusillus (Willd. ex Bernh.) Bernh. [= Cheilanthes pteridioides (Reich.) C. Chr.]. Th is species was designated the lectotype of Allosorus Bernh. by Pichi-Sermolli (1953), a choice subsequently validated by the ICBN when Cheilanthes was conserved over Allosorus (Appendix II of the Montreal Code, Stafl eu et al. 1961). Th e only phylogenetic study published to date that includes the type species of both Allosorus and Cheilanthes is that of Eiserhardt et al. (2011). In that analysis, it is unclear whether the divergence between C. maderensis (= C. pteridioides ; see Nardi and Reichstein 1985;Rothfels et al. 2012) and C. micropteris is suffi cient to justify the recognition of two genera. Th e two taxa appear in distinct, well-supported clades (clade A vs. clade C in fi g. 2B of Eiserhardt et al. 2011), but deeper relationships are poorly resolved and both clearly belong to the rapidly diversifying hemionitid lineage (clade H). Th e unequivocal assignment of Allosorus to the hemionitids by Eiserhardt et al. (2011) does, however, prevent the application of this generic name to the myriopterid clade. Any attempt to expand Allosorus to include myriopterids would encompass both Cheilanthes (conserved over Allosorus ) and Hemionitis (which has priority over both).
One potentially viable option for generic placement of the myriopterid clade would be to include it within a revised circumscription of Pellaea Link. All recent phylogenetic studies with adequate sampling of the two groups (e.g., Kirkpatrick 2007;Rothfels et al. 2008;Eiserhardt et al. 2011) strongly support the position of myriopterids as the sister group of the pellaeid clade, which includes Pellaea atropurpurea , the lectotype of the oldest generic name applicable to that clade. We are not in favor of expanding the defi nition of Pellaea to encompass the myriopterids for a variety of reasons. First, the two are quite distinct, both in terms of phylogenetic divergence and morphology. Th e myriopterids have substantially smaller ultimate segments, pubescent and/or scaly (vs. mostly glabrous) leaf blades, and sporangia that are confi ned to vein tips (vs. distributed along the veins near the segment margins). Because of these diff erences, the two groups generally have not been considered closely related, and most myriopterids would require new combinations in Pellaea . Adding to this nomenclatural upheaval is the fact that other well-defi ned genera, including Argyrochosma (Sigel et al. 2011) and Astrolepis (Beck et al. 2010), would be subsumed within such a circumscription of Pellaea , which would require additional new combinations and serve only to further undermine the distinctions among the major genera of cheilanthoid ferns.
If the expansion of Pellaea is ruled out, there remain three other generic names typifi ed by species belonging to the myriopterid clade: 1) Myriopteris , described by Fée (1852) and typifi ed by M. marsupianthes Fée; 2) Cheilosoria , named by Trevisan (1877) and lectotypifi ed by Copeland (1947) based on C. allosuroides (Mett.) Trev.; and 3) Pomataphytum , published by Jones (1930) and typifi ed by P. pocillatum M. E. Jones ( = M. lendigera ). Phylogenetic reconstructions (Grusz et al. in review) confi rm that the type species of Myriopteris and Pomataphytum fall within a single, wellsupported clade. In fact, the diploid species M. marsupianthes is thought to be one of the parents of sexual tetraploid M. lendigera (see Mickel and Smith 2004). Th us, the generic name Pomataphytum is appropriately considered a taxonomic synonym of the earlier described Myriopteris and can be eliminated as a potential name for the myriopterid clade. Copeland's (1947) lectotype of Cheilosoria belongs to the well-supported and morphologically distinctive alabamensis clade that diverges earlier in the myriopterid phylogeny (Grusz et al. in review), and the name Cheilosoria could be used for this particular group if the myriopterids were subdivided into two or more genera. However, Myriopteris predates Cheilosoria by 25 years and, when these species are assigned to a single genus (our preferred approach), Myriopteris is the correct generic name for the inclusive myriopterid clade.

Historical use of the name Myriopteris
Th e original concept of Myriopteris (Fée 1852) included 11 species, these split between two sections ( Eumyriopteris and Cheilanthastrum ) distinguished by the presence or absence of a well-developed, inframarginal false indusium. Th e Latin and French descriptions of the genus are only partly overlapping; shared elements include the highly divided fronds, the small, orbicular ultimate segments with recurved margins ("formant un bourrelet très-contracté"), and a tendency to be covered by hairs and/or scales. Myriopteris was accepted and signifi cantly expanded by J. Smith (1875: 280) who stated "the genus consists of about 20 species, distinguished from Notholaena and Cheilanthes by their small, concave, lenticular segments." Th e segregation of Myriopteris from Cheilanthes was, however, rejected by most subsequent authors (e.g., Christensen 1906;Copeland 1947;Lellinger 1965;Tryon and Tryon 1982;Kramer et al. 1990), with two notable exceptions. Pichi-Sermolli (1977) advocated a narrowed circumscription of the genus, including only the two species with prominent false indusia, viz., M. marsupianthes and M. lendigera . As shown by Grusz et al. (in review), this defi nition of Myriopteris is phylogenetically indefensible because it excludes M. mexicana , the apparent maternal progenitor of allotetraploid M. lendigera . About the same time Pichi-Sermolli was narrowing the defi nition of Myriopteris , Löve and Löve (1977) expanded it slightly by proposing a new combination for the species known as Cheilanthes covillei Maxon. Th is was done without explanation, though almost certainly refl ects the fact that this species has the small, beadlike ultimate segments emphasized in earlier circumscriptions of the genus.
Although this "microphyllous" leaf morphology is common within Myriopteris , it does not characterize the entire clade (Grusz et al. in review) and has evolved independently in other cheilanthoid lineages. Th us, the possession of small, bead-like ultimate segments does not constitute a synapomorphy for the genus as defi ned herein. In fact, our list of excluded names (see Taxonomic Treatment) includes seven taxa with bead-like segments previously ascribed to Myriopteris but more closely related to Cheilanthes s.s. (Windham et al. unpublished). Because all morphological characters used by previous authors to defi ne Myriopteris are subject to strong, positive selection in xeric-adapted cheilanthoid lineages (Hevly 1963), it is not surprising that none of them uniquely defi ne the genus. Th e totality of evidence, however, indicates that the myriopterids represent a deeply divergent clade that cannot reasonably be combined with any other in a single genus. Th erefore, we propose to resurrect Myriopteris and recircumscribe it to encompass the entirety of this well supported cheilanthoid lineage. Ideally, morphological and/or cytological synapomorphies would substantiate phylogenetic relationships inferred from DNA sequence data. However, easily observed synapomorphies distinguishing the various clades of cheilanthoid ferns are few, and homoplastic characters abound. To paraphrase Sir William Hooker (1852: 75), "Vain is the attempt to form a defi nite character which shall decide the limits of [ Cheilanthes ]," a statement that applies equally well to Myriopteris . Highly divided (decompound) leaf blades with small ultimate segments are scattered across the cheilanthoid tree and, indeed, are characteristic of ferns in general, and an indument of hairs and/or scales is one common strategy among plants used to reduce water loss in xeric habitats (Hevly 1963). Other characters useful for species-level identifi cation within myriopterids, such as vernation, are, without exception, shared with other distantly related cheilanthoid ferns.
Molecular analyses spanning the diversity of cheilanthoid species (Windham et al. unpublished) illuminate one particularly useful character distinguishing Myriopteris , as defi ned herein, from Cheilanthes s.s. Th e taxa most closely related to the type species of the latter [ C. micropteris plus all Australian Cheilanthes and a group of South American species including the C. scariosa (Sw.) C. Presl complex of Tryon and Tryon (1982), C. obducta Mett. ex Kuhn, and C. fractifera R. M. Tryon] have 32 small spores per sporangium when sexual, and 16 large spores per sporangium when apomictic. Th is intriguing cytological synapomorphy results from the elimination of a premeiotic mitosis in the cell lineages generating the sporocytes (Windham et al. unpublished). Aside from a few species of the distantly related genus Notholaena , all other cheilanthoid ferns so far examined (including every Myriopteris species; Grusz et al. in review) produce 64 small spores per sporangium in sexual individuals and 32 large spores per sporangium in apomicts. Th is character appears to provide an absolute separation between Myriopteris and Cheilanthes s.s. , and is easily observed using a dissecting microscope. In combination with diff erences in spore ornamentation (see Tryon and Lugardon 1991), leaf venation , and geographic distribution, this feature provides a clear distinction between the two genera. For diagnostic purposes, then, Myriopteris Fée emend. Grusz & Windham diff ers from Cheilanthes s.s. (i.e., C. micropteris and its close relatives) in its production of 64 small or 32 large (vs. 32 small or 16 large) spores per sporangium; mostly cristate or rugulose (vs. echinate, granulose, or verrucate) spore ornamentation; a lack of obvious vein endings near the margins of the ultimate segments (vs. often prominent hydathodes), and a largely North and Central American (vs. exclusively South American/Old World) distribution.
Distribution. Species of Myriopteris range from southern Canada through the Caribbean and Central America to southern Chile, with one species ( M. rawsonii ) endemic to Namibia and South Africa. Mexico is the center of species diversity for the genus; 34 of the 44 species can be found in Mexico, and seven of these are endemic.