A fossil Osmunda from the Jurassic of Sweden—reconciling molecular and fossil evidence in the phylogeny of Osmundaceae

The systematic classification of Osmundaceae has long remained controversial. Recent molecular data indicate that Osmunda is paraphyletic, and needs to be separated into Osmundastrum and Osmunda s. str. Here we describe an exquisitely preserved Jurassic Osmunda rhizome (O. pulchella sp. nov.) that combines diagnostic features of Osmundastrum and Osmunda, calling molecular evidence for paraphyly into question. We assembled a new morphological matrix based on rhizome anatomy, and used network analyses to establish phylogenetic relationships between fossil and extant members of modern Osmundaceae. We re-analysed the original molecular data to evaluate root-placement support. Finally, we integrated morphological and molecular data-sets using the evolutionary placement algorithm. Osmunda pulchella and five additional, newly identified Jurassic Osmunda species show anatomical character suites intermediate between Osmundastrum and Osmunda. Molecular evidence for paraphyly is ambiguous: a previously unrecognized signal from spacer sequences favours an alternative root placement that would resolve Osmunda s.l. as monophyletic. Our evolutionary placement analysis identifies fossil species as ancestral members of modern genera and subgenera. Altogether, the seemingly conflicting evidence from morphological, anatomical, molecular, and palaeontological data can be elegantly reconciled under the assumption that Osmunda is indeed monophyletic; the recently proposed root-placement in Osmundaceae—based solely on molecular data—likely results from un- or misinformative out-group signals.


INTRODUCTION 1
The royal ferns (Osmundales) comprise about 20 extant species currently classified in four 2 genera, i.e. Osmunda L., Osmundastrum C.Presl, Leptopteris C.Presl, and Todea Bernh. This 3 small group of ferns is remarkable in many respects and, consequently, has attracted 4 considerable scholarly attention. Its members represent the most primitive of all 5 leptosporangiate ferns (e.g. Pryer et al., 2004;Smith et al., 2006Smith et al., , 2008Schuettpelz & Pryer, 6 2007), with features that have been interpreted to be intermediate between Eusporangiatae and7 Leptosporangiatae (e.g. Bower, 1891, 1926;Tidwell & Ash, 1994). Detailed investigations of 8 their anatomy (e.g. Faull, 1901(e.g. Faull, , 1909Seward & Ford, 1903;Hewitson, 1962), cytology and 9 genetic structure (e.g. Strasburger, 1900;Yamanouchi, 1910;Digby, 1919;Sharp, 1920;10 Manton, 193910 Manton, , 1945Manton & Smiles, 1943;Tatuno & Yoshida, 1966, 1967Klekowski, 11 1970Klekowski, 11 , 1973Yatabe et al., 2009), and evolution (e.g. Kidston & Gwynne-Vaughan, 1907-12 1910, 1914Miller, 1967Miller, , 1971Yatabe, Nishida & Murakami, 1999;Metzgar et al., 2008;13 Escapa & Cúneo, 2012) render the Osmundales one of the most intensively studied groups of 14 ferns. Moreover, in contrast to their rather limited diversity today, Osmundales have a 15 uniquely rich and diverse fossil record (e.g. Arnold, 1964;Miller, 1971) currently considered 16 to include more than 150 species, over 25 genera, and at least three families (e.g. Tidwell & 17 Ash, 1994;Tian, Wang & Jiang, 2008;Wang et al., 2014). This extensive fossil record has 18 been reviewed in several key works (Miller, 1971;Tidwell & Ash, 1994;Tian et al., 2008;19 Wang et al., 2014). 20 The monophyly of Osmundales and their isolated position as the first diverging lineage 21 within leptosporangiate ferns are firmly established (see, e.g. Hasebe et al., 1995;Schneider 22 et al., 2004;Pryer et al., 2004;Smith et al., 2008). However, the resolution of systematic 23 diameter and entirely parenchymatous (Fig. 2). A thin region at the outermost periphery of the 1 pith consists of a few rows of parenchyma cells that are considerably more slender (ca 20-30 2 µm wide) than those in the central portion of the pith (usually ≥ 50 µm wide) (Figs 2B,3A,B,3 H); furthermore, cell walls in some regions of the pith periphery may be thicker and more 4 clearly visible than in the centre (Figs 2A, 3B). However, there is no evidence for the 5 presence of an internal endodermis or internal phloem. Given that endodermal layers are 6 recognisable in the stem and petiole cortices (e.g. Fig. 4F), we are positive that the absence of 7 an internal endodermis is an original feature, and not the result of inadequate preservation. 8 The xylem cylinder is ca 0.4 mm and ca 8-12 tracheids thick, and dissected by narrow, 9 mostly complete, immediate leaf gaps into about 20 xylem segments in a given transverse 10 section. The phloem forms an entire ring around the stele; it is most easily recognisable 11 opposite a leaf gap, where it forms a narrow wedge-shaped patch of large, thin-walled cells 12 that projects slightly towards the gap in transverse section (Figs 2A, 3A). 13 The cortex of the stem is bi-layered (Figs 1E, 2, 6A). The inner layer is ca 0.5-0.8 mm 14 thick, consists entirely of parenchyma, and contains about ten leaf traces in a given transverse 15 section ( Fig. 2A). The outer cortex is considerably thicker (ca 1.5-2.5 mm thick), and consists 16 entirely of homogeneous sclerenchymatic tissue (Figs 1E,2). Abundant leaf traces (about 20 17 in a given transverse section; e.g. Fig. 2A) and rootlets traversing the outer cortex (Figs 1C,18 D,2) appear to have altered the original orientation of the sclereids, resulting in a somewhat 19 patchy appearance of the outer cortical tissue (Fig. 2). 20 Phyllotaxy of the stem is helical with apparent contact parastichies of 8 and 13 (Fig. 1B, 21 E). Leaf-trace formation begins with the appearance of a single protoxylem strand in an 22 eccentric position (about two-thirds to three-quarters distance from the pith; Fig. 3A) in a 23 stelar metaxylem segment. Distally, the protoxylem becomes associated with an increasing amount of parenchyma on its adaxial side (making it effectively endarch for the rest of its 1 course), first occupying only the centre of the segment (resulting in an O-shaped xylem 2 segment), then connecting with the pith (resulting in a U-shaped xylem segment), and 3 ultimately forming the usually complete, narrow leaf gap with the departure of the trace. 4 Departing leaf traces are oblong, only slightly curved adaxially, ca 300-350 µm wide and two 5 to four tracheids (ca 80-100 µm) thick (Figs 2, 3D, E), and diverge from the axis at angles of 6 ca 20-40° (Figs 1D, 2B). 7 In its course through the stem, a leaf-trace vascular bundle becomes enveloped by 8 increasing layers of tissue through which it successively passes: first by phloem and 9 endodermis from the stele upon entering the inner cortex; by a sheath of parenchyma from the 10 inner cortex as it enters the outer cortex (Fig. 2); and finally by a cylindrical sclerenchyma 11 sheath from the outer cortex as it departs from the stem (Fig. 1E). The initial bifurcation of the 12 leaf-trace protoxylem occurs in the outermost portion of the cortex or in the petiole base (Fig.  13 3F, G). 14 In the inner cortex of the petiole, thick-walled fibres appear in form of a small irregular 15 mass adaxial to the vascular bundle (Fig. 4C,D). This develops distally into a thick band 16 lining the bundle concavity (Figs 4E,5A,B), and may further differentiate into two lateral 17 masses connected only by a rather thin strip (Fig. 4F, G). Apart from the sclerenchyma inside 18 the vascular-bundle concavity, the inner cortex of the petiole consists entirely of parenchyma. 19 The sclerenchyma cylinder of the petiole has an even thickness that increases from 20 about 300 µm near the petiole base to ca 500 µm distally. Its composition is heterogeneous: 21 near the petiole base, it contains a crescentic, abaxial arch of particularly thick-walled fibres 22 (Figs 1E,4,5,6B); distally, this arch begins to develop two lateral masses 6B) and ultimately two lateral masses and one abaxial arch of thick-walled fibres whose lumina 1 are more-or-less entirely occluded (Figs 4G, 6B). 2 The petiole bases are flanked by a pair of stipular wings that consist initially of 3 parenchyma only; as the wings grow wider in more distal portions, they develop a patch of 4 thick-walled fibres (Figs 4D, 6B) that forms an entire, elongate strip (Figs 4F, G, 6B). The 5 parenchymatic ground tissue of the stipular wings is well-preserved only in the innermost 6 regions of the mantle (Figs 1B, C, E); outwards, it appears to be either increasingly degraded 7 or to have been removed by the abundant penetrating rootlets. In the outermost portions of the 8 mantle, all that remains of the stipular wings are usually just the isolated, elongate strips of 9 thick-walled fibres interspersed between petioles and rootlets (Fig. 4G). 10 Each leaf trace is usually associated with a single rootlet that diverges laterally at the 11 point of departure from the stele. The rootlets typically measure about 0.5 mm in diameter, 12 contain a diarch vascular bundle, parenchymatic ground tissue with interspersed 13 sclerenchymatic fibres, and a sclerenchymatic outer cortical layer.  Bomfleur et al., in press). Some parenchyma cells, especially those 20 adjacent to xylem bundles in roots and leaf traces, contain varying amounts of discrete, 21 smooth-walled, spherical or oblate particles ca 1-5 µm in diameter that have been interpreted 22 as putative amyloplasts (Bomfleur et al., in press). Cell nuclei measure ca 10 µm in diameter, and contain nucleoli and chromosomes . Chromatid strands have a diameter of 1 0.3-0.4 µm (Fig. 3O). 2

Phylogenetic analyses 3
Phylogenetic relationships among fossil and modern members of the Osmundaceae based on 4 rhizome anatomy (Fig. 7). Especially remarkable is the diversification of subgenus Osmundastrum as revealed by 1 our independent coding of the individual fossil records from Neogene (Miller, 1967;2 Matsumoto & Nishida, 2003), Paleogene (Miller, 1967), and Cretaceous deposits (Serbet & 3 Rothwell, 1999); the individually coded fossil and extant representatives assigned to O. 4 cinnamomea show greater morphological disparity than expressed between the separate 5 species of any other subgenus and genus. 6 Merging fossil and extant taxa into a molecular backbone topology (Fig. 8). -Of all taxa 7 placed via EPA (evolutionary placement algorithm), Osmunda pulchella is the species that is 8 most incongruently placed between the different weighting schemes: Using parsimony-based 9 character weights, the EPA places Osmunda pulchella at the root of Claytosmunda, whereas it 10 is placed either between Osmundastrum and the remaining Osmunda s. str. or at the root of 11 the Plenasium clade using model-based character weights. Single position swaps also occur in 12 most of the other Jurassic species [O.  Re-visitation of the outgroup-inferred Osmundaceae root (Fig. 9). -The gene jackknifing 4 and single-gene analyses reveal ambiguity concerning the position of the Osmundaceae root 5 in the data of Metzgar et al. (2008). As in the original analysis, support for backbone branches 6 is effectively unambiguous based on the concatenated data, and places the outgroup between 7 Osmundastrum and the remainder of the family, resolving the traditional genus Osmunda 8 (Osmunda s.l.) as a grade ('paraphyletic Osmunda scenario'). The signal for this root 9 placement stems from the two coding plastid gene regions (atpA and rbcL). In the more (but 10 not most) variable spacer regions (rbcL-accD, atpB-rbcL, and trnL-trnF to a lesser degree), 11 however, a competing signal is found resolving genus Osmunda s.l. as a clade ('monophyletic 12 Osmunda scenario'). The most variable non-coding spacer regions (trnG-trnR; rps4-trnS; and 13 trnL-trnF to some degree) provided only ambiguous signals including potential outgroup-14 branch placements deep within the Leptopteris-Todea and Osmunda sub-trees and showed a 15 preference for an Osmundastrum-Leptopteris-Todea clade as sister to Osmunda s. str. 16 The gene-jackknifing results showed that the exclusion of either one or both coding 17 regions (atpA, rbcL)-which together account for 33% of distinct alignment patterns in the 18 concatenated matrix-decreased support for the split leading to an Osmunda grade with 19 Osmundastrum resolved as sister to the remainder of the family, whereas the support for the 20 alternative of an Osmunda clade or an Osmundastrum-Leptopteris-Todea clade was increased. Osmundaceae in the regional fossil flora 3 Osmunda pulchella sp. nov. is among the earliest fossil Osmunda rhizomes known so far, and 4 the first such find from the Mesozoic of Europe. Whole plants are rarely fossilized, so 5 identification of fossils depends on recognizing diagnostic characters in various dispersed 6 organs. Moreover, some isolated organs can only be identified to taxa under special 7 preservational states (e.g., where anatomical details are retained). Fossil evidence for 8 Osmundaceae occurs in three main forms: (1) permineralized axes with vascular, cortical and 9 petiolar anatomy characteristic of the family; (2) compressions and impressions of foliage 10 (either fertile or sterile); (3) dispersed spores with sculptural characters typical of fertile 11 macrofossil or extant representatives of the family. 12 Permineralized osmundaceous axes have a long-ranging and geographically broad fossil 13 record extending back to at least the Permian of both hemispheres (Gould, 1970;Tian et al., 14 2008;Wang et al., 2014). These fossils are highly informative of the anatomical evolution of 15 the group since they preserve the three-dimensional architecture of axial tissues and the 16 surrounding sheath of petioles (Miller, 1971). They provide further information on 17 osmundacean ecology, since the excavations or coprolites of various invertebrates are 18 commonly preserved within the cortical tissues or petiole sheath (Tidwell & Clifford, 1995). 19 However, the occurrences of permineralized axes are generally restricted to sedimentary rocks 20 with a high proportion of volcanogenic components. Free silica and, in some cases, carbonate 21 ions are liberated in particularly high concentrations from the breakdown of glass and 22 unstable calc-silicate minerals, especially in sediments derived from mafic to intermediate volcanic terrains (Jefferson, 1987). These ions preferentially link to free hydrogen bonds of 1 holocellulosic complexes in buried plant matter, entombing the original cell walls in opaline 2 silica, quartz, or calcite. The exceptional circumstances of such preservational conditions 3 mean that permineralized osmundaceous stems have a patchy record [see Tidwell (2002)  fronds are typically assigned to Cladophlebis Brongn., although not all forms referred to this 16 fossil genus are necessarily osmundacean. Collectively, the record of fossil osmundacean 17 foliage matches that of the rhizomes, extending from the Permian to Cenozoic and being 18 distributed on all continents (Herbst, 1971;Miller, 1971;Anderson & Anderson, 1985;Hill et 19 al, 1999;Collinson, 2001). Foliage referable to Todites or Cladophlebis is widespread in the 20 Mesozoic of Europe and is extensively represented in Rhaetian to Early Jurassic strata of 21 southern Sweden (Nathorst 1878;Antevs, 1919;Johansson, 1922;Lundblad, 1950;Pott & 22 McLoughlin, 2011). 23 Spores attributed to Osmundaceae found in situ within fossil sporangia or dispersed within 1 sediments are spherical to triangular and typically bear irregularly arranged grana, bacula or 2 pila of variable form and size. More rarely, the spore surface is scabrate or laevigate. When Baculatisporites species are common elements of palynofloras recovered from the uppermost 14 Triassic to Middle Jurassic strata of Sweden (Tralau, 1968;Lund, 1977;Guy-Ohlson, 1986;15 Lindström and Erlström, 2006;Larsson, 2009), indicating that the family had an important 16 role in the ecology of the herbaceous stratum of the regional mid-Mesozoic vegetation. 17 Osmundaceae underwent a notable decline in both relative diversity and abundance 18 accompanying the rise of the angiosperms in the Cretaceous (Nagalingum et al., 2002;19 Coiffard, Gomez & Thevenard, 2007) and this trend appears to have persisted through the 20 Cenozoic resulting in the family's low representation and, for some genera, relictual 21 distribution today (Collinson, 2001). 22

Systematic placement of fossil Osmunda rhizomes among modern Osmundaceae 19
Phylogenetic network analysis. The placement of the other fossil taxa is overall in accordance with the basic assumption 11 that they should be less derived-and thus placed closer to the centre of the network-than 12 their extant relatives. However, there is one major exception: O. dowkeri from the Paleogene 13 is the furthest-diverging (i.e. most derived) of all fossil and extant species in the Plenasium 14 group. This relates to its unusually complex stele organization, which contains by far the 15 largest number of xylem segments of all species analysed (exceeding 30, compared to less 16 than 12 in all other Plenasium and less than 20 in most other Osmunda. 17 Notably, a subdivision into two putatively monophyletic subgenera Osmunda sensu 18 Yatabe et al. and Claytosmunda generates two taxa without discriminating anatomical and 19 morphological features (potential aut-or synapomorphies according to Hennig, 1950). 20 Miller's paraphyletic subgenus Osmunda accommodates the fossil taxa, whereas the concept 21 of Yatabe et al. (1999,2005) precludes infrageneric classification of most fossil species (Fig.  22   7).
Compatibility with vegetative morphology. -The systematic relationships revealed from our 1 analysis of anatomical characters of the rhizomes reflect the distribution of gross 2 morphological and fertile features within Osmundaceae very well. The isolated position and 3 tight clustering of subgenus Plenasium, for instance, finds support through morphological 4 data in the form of its invariant, unique frond morphology: unlike any other modern 5 Osmundaceae, all extant Plenasium species are characterized by having invariably simple-6 pinnate and hemi-dimorphic fronds. Also the rather wide dispersion of the (paraphyletic) 7 subgenus Osmunda Miller is congruent with the variable frond morphology and dimorphism 8 in this group, ranging from pinnate-pinnatifid [e.g. O. claytoniana (similar to O. 9 cinnamomea)] to fully bipinnate and from fully to variably hemi-dimorphic. 10 The only major topology where anatomical data alone probably fail to generate a 11 realistic divergence distance occurs in the branch including Todea and Leptopteris. These 12 genera, with their rhizome anatomy being overall similar to those of Osmunda and especially 13 Osmundastrum (Hewitson, 1962; but see Fig. 7 which we interpret to result in "least conflicting" placements at varying root positions; the 22 EPA is designed to optimize the position of a query taxon within a pre-defined backbone 23 topology. Since O. pulchella, and other fossil taxa, show character combinations of genetically distant taxa, the model-based weights in particular will down-weigh the relevant 1 characters. Maximum parsimony has a much more naïve approach in this respect, which may 2 be beneficial for a plausible placement of the fossils. Nevertheless, the fact that this down-3 weighting results in a placement close to the roots, but not in the tips of sub-trees, indicates 4 that the remaining character suite is plesiomorphic in general, thus supporting the 5 interpretation of fossil taxa such as O. pulchella as ancestors of extant clades and possibly 6 species (Figs 7 and 8). specialization in the subgenus, which is supposed to have reached its heyday in 20 distribution and diversity during the Paleogene (Miller, 1971). 21 (5) All fossil Osmundastrum can be unambiguously identified as such, despite the wide 22 stratigraphic age-span (Cretaceous, Paleogene, and Neogene) and 'trans-Pacific' 23 geographic distribution; it is interesting to note, however, that the rhizomes of O.
cinnamomea show a far greater disparity in anatomical characters than all other 1 subgenera and even genera of modern Osmundaceae, indicating the existence of probably 2 more than just one Osmundastrum species in the past (Fig. 7). (7) the Early Cretaceous Todea tidwellii may be as related to modern Leptopteris as it is to 7 Todea. 8

Re-evaluation of the Generic Status of Osmundastrum 9
The intermediate character combination and the resulting systematic placement of Osmunda 10 pulchella and other Jurassic species between Osmundastrum and subgenus Osmunda Miller as 11 detailed above is incompatible with the current treatment of Osmundastrum as a separate 12 genus. In the following section we, therefore, provide a detailed re-evaluation of the sum of 13 evidence that has been used to invoke generic separation of Osmundastrum. We begin with 14 what is perhaps considered the most novel and reliable body of evidence-molecular data-15 and continue with additional evidence from morphological, anatomical, and hybridization 16 studies. 17 Molecular data (Fig. 9). -The comprehensive multi-locus phylogeny of Metzgar et al. 18 (2008) has recently been interpreted to fully support a separate generic status of 19 Osmundastrum as suggested by Yatabe et al. (1999). Indeed, inter-generic and inter-20 subgeneric relationships based on the molecular matrix used by Metzgar et al. (2008; reproduced here in Fig. 9) receive nearly unambiguous support from the concatenated gene 1 matrix. 2 However, our root-stability analysis revealed that the inferred paraphyletic status of 3 Osmunda s.l. is not unambiguously supported by all gene regions (Fig. 9). Although receiving 4 strong support from the two coding regions (rbcL-gene, atpA-gene), the molecular data 5 matrix of Metzgar et al. (2008) also yields a strong conflicting signal from three relatively 6 conserved spacer sequences (i.e. trnL-trnF, atpB-rbcL, and rbcL-accD) that indicate an 7 alternative root placement between Leptopteris-Todea and the remaining Osmunda s.l. -8 offering an equally valid interpretation that would resolve Osmunda s.l. as monophyletic. 9 The root-placement problem may be partly due to the incomprehensive selection of out-10 group taxa, which is limited to four samples of leptosporangiate ferns in the matrix of 11 include five other extant orders apart from Gleicheniales (see, e.g., Pryer et al., 2004;Smith et 16 al., 2006;Schuettpelz & Pryer, 2007, 2008. In order to obtain a more informative signal, a 17 comprehensive outgroup selection should include taxa from the sister clades of the 18 Polypodiopsida (Marattiopsida and Equisetopsida) and all major lineages within the 19 Polypodiopsida, in particular Hymenophyllales and Schizaeales. Since the Gleicheniales are 20 relatively derived in comparison to the Osmundales, their members may inflict outgroup long-21 branch attraction with Osmundastrum [see Figs S1 (note the long terminal edge bundles) and 22

S2 in ESA]. 23
Anatomy. -Rhizomes of extant O. cinnamomea exhibit a few peculiar and supposedly 1 unique characters, including (1) the common occurrence of an internal endodermis; (2) the 2 rare occurrence of a dissected, ectophloic to amphiphloic condition of the stele; (3) the 3 protoxylem bundle bifurcating only as the leaf trace enters into the petiole base; (4) the 4 sclerenchyma ring of a petiole base containing one abaxial and two lateral masses of thick-5 walled fibres; (5) roots arising from the leaf traces usually singly, and only rarely in pairs; and 6 (6) a patch of sclerenchyma adaxial to each leaf trace in the inner cortex (e.g. Hewitson, 1962;7 Miller, 1971). 8 The first two characters occur only inconsistently, and are notably absent in Cretaceous 9 to Neogene fossil representatives of Osmundastrum (Miller, 1971;Serbet & Rothwell, 1999), 10 indicating that these characters might represent recently acquired traits (Miller, 1971). 11 Moreover, the dissected stele condition, with both endoderms connecting through a leaf gap, 12 occurs only very rarely below incipient rhizome bifurcations, and could only be revealed after 13 thorough investigations of serial sections of over one hundred specimens (Faull, 1901(Faull, , 190914 Hewitson, 1962). The significance of both characters as diagnostic criteria is thus 15 questionable. 16 The point of protoxylem bifurcation and the arrangement of patches of thick-walled 17 fibres in the petiole sclerenchyma ring occur consistently, and arguably form appropriate 18 diagnostic criteria for Osmundastrum. However, among the remaining Osmunda s.l. species, 19 these same two characters are treated as diagnostic features of only specific or subgeneric 20 rank (Miller, 1971); it would thus be inconsistent to weight them more strongly in the 21 delimitation of Osmundastrum only. 22 Roots arising in most cases singly, as opposed to mostly in pairs in the remaining 23 Osmunda, is a useful distinguishing character of Osmundastrum and Osmunda pulchella, although this feature is only inconsistent and may be difficult to observe (Hewitson, 1962;1 Miller, 1971). The occurrence of sclerenchyma patches adaxial to the leaf traces in the inner 2 stem cortex is the only invariant and unique distinguishing character of Osmundastrum that 3 we consider might warrant separation above species level. Apart from Osmundastrum, this 4 feature is found only in Todea and not in its sister genus Leptopteris (Miller, 1971). 5 Morphology. -Morphological features that are commonly cited as diagnostic of 6 Osmundastrum include (1) the usually complete frond dimorphism; (2) pinnate-pinnatifid 7 frond architecture; and (3) abaxial hair cover on pinna rachides (Metzgar et al., 2008). 8 The use of the type of frond architecture and dimorphism as a strict diagnostic character has 9 been shown to be problematic (e.g. Hewitson, 1962). claytoniana. In addition, some common varieties and growth forms of O. cinnamomea 12 produce only hemi-dimorphic fronds (e.g. Torrey, 1840;Britton, 1890;Kittredge, 1925;13 Steeves, 1959;Werth, Haskins & Hulburt, 1985), with some having apical fertile portions 14 resembling those of O. regalis (see, e.g. Hollick, 1882;Murrill, 1925;Werth et al., 1985) and 15 others having intermittent fertile portions like those of O. claytoniana (see, e.g. Day, 1886;16 Werth et al., 1985). Moreover, completely dimorphic fronds occur also predominantly in O. 17 lancea, regularly in O. japonica, and sporadically in O. regalis (Hooker & Baker, 1883;18 Chrysler, 1926;see Hewitson, 1962). Notably, such ranges of variation are encountered only 19 in the species complex including Osmundastrum and Osmunda subgenus Osmunda Miller (= 20 subgenera Claytosmunda and Osmunda Yatabe et al.). 21 Finally, fronds of all Osmunda s.l. species emerge with a more-or-less dense abaxial 22 indumentum and differ merely in the duration to which the trichome cover is retained in the 23 course of frond maturation (Hewitson, 1962). In fully mature fronds of all species considered, most of the hair cover is ultimately lost, with O. cinnamomea [especially O. cinnamomea var. 1 glandulosa Waters (see Waters, 1902;McAvoy, 2011)] merely tending to retain greater 2 amounts of hairs than O. claytoniana, and those in turn more than other species (Hewitson, 3 1962). In summary, we follow Hewitson (1962) and consider none of these morphological 4 features to provide consistent and reliable diagnostic characters for separating Osmundastrum 5 from subgenus Osmunda Miller. 6 Hybridization. -Metzgar et al. (2008, p. 34) suggested that the existence of hybrids can be 7 used to decide about the elevation of subgenera to generic ranks. A range of natural hybrids, 8 intra-and inter-subgeneric, are known to occur in Osmunda s. str.: O. × ruggii R.M.Tryon in 9 eastern North America (O. regalis × O. claytoniana;Tryon, 1940;Wagner et al., 1978), O. × 10 mildei C.Chr. in southern China (O. japonica × O. vachellii Hook.;Zhang et al., 2008;Kato 11 et al., 2009) (Kato, 2009;Yatabe et al., 2009;Tsutsumi et al., 2012). The seeming 15 absence of naturally occurring hybrids involving Osmundastrum has been interpreted to result 16 from its particularly isolated position within Osmunda s.l. (Miller, 1967(Miller, , 1971 Leptopteris-Todea and Osmunda s.l.
In our opinion, this latter option integrates the seemingly conflicting evidence from 1 studies of the morphology, anatomy, molecular data, and fossil record of Osmundaceae in a 2 much more realistic and elegant way, and-beyond that-offers a more practical taxonomic 3 solution. We, therefore, argue that