Todea from the Lower Cretaceous of western North America: implications for the phylogeny, systematics, and evolution of modern Osmundaceae^†

Specific diagnosis

Small osmundaceous trunks with closely spaced frond bases and adventitious roots. Rhizomes ∼4.5 cm long, up to 3 cm wide; branching distally; phyllotaxis helical; with heterogeneous sclerotic cortex. Cortex with indistinct outer boundary and prominent sclerotic ring surrounding frond and root traces from inner margin of sclerotic cortex. Rhizome dictyoxylic and ectophloic, with up to five cauline bundles and thin C-shaped leaf traces; one foliar protoxylem at divergence of trace, dividing in two in inner cortex, and increasing to six to eight in stipe base. Pith sclerenchymatous, up to 2 mm in maximum diameter; sclerenchyma lobes of pith diverge as a rod adaxial to each leaf trace becoming C-shaped at stipe base. Sclerotic ring of stipe heterogeneous; thinner adaxially. Diarch adventitious roots diverging from cauline bundles, one per frond.

Stem

The stem increases in diameter from less than 2 mm at the most proximal level preserved to more than 2 cm distally. The stem divides ∼2 cm from the base of the specimen, forming two steles distally (Fig. 8, at arrows). Although each branch is surrounded by diverging frond bases, both remain within a common sclerotic cortex for the length of the specimen preserved.

The stele is characterized by a prominent sclerotic pith (Figs. 2, 3, 6). At some levels cells of the pith are thick walled throughout (Figs. 3, 6), but at other levels there are thin-walled cells at the center (Fig. 2). Distal to the level of branching, cells are not preserved at the center of the pith (Fig. 8), suggesting that they were thin walled in life or immature at the time of fossilization. In cross sections the pith region is stellate (Figs. 2, 3, 6), a shape that results from large sclerotic bundles separating from the pith and diverging to the adaxial side of each leaf trace (Figs. 4, 6). In a series of sections through which a frond trace diverges, the pith first extends radially forming ribs and then each rib separates as a rod-shaped bundle near the adaxial surface of the frond trace. As each trace diverges through the cortex of the rhizome, the rod of sclerenchyma first becomes continuous with the sclerotic ring of the stipe (Fig. 6, at right), and then detaches from the ring to occupy the space within the C-shaped xylem bundle (Figs. 5, 7). Through the same distance the rod of sclerenchyma diminishes in radial thickness until it becomes a thin zone that lies adjacent to the C-shaped frond trace (Fig. 5, 7).

The stem stele consists of a small number of cauline bundles and diverging leaf traces throughout the length of the specimen. At the base, there are only three cauline bundles (Fig. 2, at arrows) and diverging leaf traces, but the number increases distally. At the level of Fig. 2, there are three frond traces diverging from the stele in a single section and about five orthostichies of leaves. More distally at the level of Figs. 3 and 6, the number of diverging traces has increased to four, and there are about 10 orthostichies. Above the level of branching, distortion and incomplete preservation make the number of cauline bundles and phyllotaxis difficult to determine (Fig. 8).

Cauline vascular architecture and leaf trace divergence

In a typical cross section of the rhizome, the stele includes both elliptical cauline bundles and C-shaped leaf traces, which are interconnected in various combinations (Fig. 6). Some leaf traces are not connected to a cauline bundle (Fig. 6, at right), while others have a cauline bundle attached either to the right or left side. A few traces appear to have a cauline bundle attached to each side (Fig. 6, at upper right). In addition, most sections show a root trace diverging from the periphery of a cauline bundle (Fig. 6, at left). These configurations are typical of osmundaceous ferns with dictyoxylic steles (11; 23).

Production of a leaf trace can be described from an acropetal series that begins at a level where a cauline bundle has no protoxylem strand (i.e., Fig. 6, at left). Progressing distally, a protoxylem strand appears in the bundle, and then the bundle becomes bilobed (Fig. 6, at upper left). The lobe with no protoxylem represents the cauline bundle, while the other lobe is the developing leaf trace. Continuing distally, the trace broadens tangentially, and metaxylem adaxial to the protoxylem strand diminishes in thickness. The leaf trace then becomes endarch and is clearly C-shaped (Fig. 6, at right). Distal to the level where the leaf trace separates from the cauline bundle, the cauline bundle fuses to an adjacent bundle/trace complex, thereby closing a leaf gap.

The single protoxylem strand located at the inner margin of the diverging leaf trace divides in two within the inner zone of the sclerotic cortex (Fig. 6, at right). The number of protoxylem strands continues to increase distally up to approximately six to eight at the level of frond divergence (Fig. 7, at arrowheads). Leaf trace metaxylem is typically one to two cell layers thick. It consists of tracheids that are 40–50 μm in diameter and that display scalariform secondary wall thickenings (Fig. 9).

Stipe bases

As is characteristic of living Todea barbara, cross sections of T. tidwellii reveal several stipes at various levels of divergence from the periphery of the rhizome (Figs. 2, 3), producing an irregular surface and obscuring the outer boundary of the cortex (Figs. 2, 3). Stipe bases are oval to elliptical in cross section and often distorted. The stipe bases are relatively isodiametric toward the outer margin of the rhizome cortex, broadening tangentially at the level of divergence (Figs. 2, 3). Individual stipe bases have a thin C-shaped xylem bundle (1–2 cells thick) with slightly recurved margins (Figs. 5, 7). There is a narrow space surrounding the xylem, that represents the position of phloem which otherwise is not preserved. The thin C-shaped sclerenchyma bundle occupies a position adjacent to the space on the concave adaxial side of the xylem (Figs. 5, 7). In some sections a layer of cells with prominent walls is preserved on the convex abaxial side of the vascular bundle (Fig. 7). Such cells are continuous with sclerenchyma of the cortex (Fig. 5), which is largely replaced by destructive calcite recrystallization. However, the preservation of a few patches of cells in this zone (Fig. 5) reveals that the cortical ring was originally composed of sclerenchyma throughout. Cortical sclerenchyma cells are 20–45 μm in diameter and have prominent lumens. There is a space between the leaf trace and the sclerenchyma ring like that which is occupied by parenchymatous ground tissue in living osmundaceous ferns (11; 1).

The sclerenchyma ring is heterogeneous, being distinctly thinner on the adaxial surface than in other areas (Figs. 2, 3, 5). It forms the outer margin of the stipe in all sections. There is no evidence of thin-walled paranchyma or lateral parenchymatous wings like those of other fossil and living osmundaceous species. Because of the calcite recrystallization of the sclerenchyma ring, we cannot be certain that lateral wings were absent in life.

Adventitious roots

As far as can be determined, one adventitious root trace diverges from the periphery of a cauline bundle in association with each frond trace (Figs. 4, 6). Root traces diverge from a cauline bundle (Fig. 6, at left), traverse the space between the stele and sclerotic cortex of the rhizome, and become surrounded by a ring of sclerenchymatous cells with small cell lumens (Fig. 6, at right). Roots extend obliquely toward the periphery of the trunk within the rhizome cortex and stipe bases (Fig. 4), eventually extending into the substrate from the abaxial surface of a stipe. There is complete histological continuity between the cortex of the rhizome and that of the root. Therefore, adventitious roots of T. tidwellii differentiated simultaneously with the other vegetative organs of the plant. Xylem of the roots is diarch, consisting of a plate of 6–8 tracheids in cross section (Figs. 3, 4).

Phylogenetic analyses

To test our hypothesis of generic affinities for T. tidwellii and of the broader phylogenetic relationships within Osmundaceae, we conducted a numerical cladistic analysis (maximum parsimony) using a combination of 1227 nucleotide sequences of the chloroplast gene rbcL (35) and 41 morphological characters (i.e., characters 1228–1268; Appendices 1, 2). The resulting tree was compared to results of previous phylogenetic analyses of the family (Figs. 10, 11). Results of this analysis yielded one most parsimonious tree of 1819 steps (CI = 54; RI = 51), that is identical to the results of 35) except that T. tidwellii appears as the sister to T. barbara (Fig. 12). In this tree (Fig. 12) O. cinnamomea (i.e., Osmundastrum) is sister to the remaining Osmundaceae, and representatives of subgenera Osmunda and Plenasium (i.e., O. regalis + O. vachellii) form the sister group to a clade that includes the species of Todea and Leptopteris (i.e., L. hymenophylloides + [T. barbara + T. tidwellii]). Bootstrap support values for nodes on the tree range from 89 to 96% (Fig. 12).

Systematics and phylogeny

The genus Todea is represented by the well-known species T. barbara and a second living species, T. papuana, which has been briefly described only once (10). Most features of T. papuana have not been characterized or critically compared to those of T. barbara and other osmundaceous ferns. The most distinctive known characters separating the two living species of Todea are differences in the sclerenchyma sheath surrounding the frond trace and sclerenchyma distribution within the stipular wings, but 10) also stresses that the latter character is highly variable at different levels with a single frond of T. papuana. Therefore, we are unsure whether more detailed studies of T. papuana will support segregation of such ferns at the species level, or if they will ultimately be regarded as falling within the range of variation that encompasses T. barbara. Because of a paucity of data, T. papuana was not included in the phylogenetic analyses.

Classical studies of osmundaceous ferns have established that living species conform to three genera Osmunda, Todea, and Leptopteris and that Osmunda s.l. may be further segregated to Osmunda L., Osmundastrum Presl., and Plenasium Presl. (6; 16; 14). Most authors recognize Todea and Leptopteris as separate genera, but others favor including species of both within Todea (e.g., 11). Whereas some authors recognize segregates of Osmunda as distinct genera, most consider them to be subgenera (11; 16; 14). Except for O. claytoniana, which Miller regards as a member of subgenus Osmunda and 14) recognizes as a species of Osmundastrum, species composition of the osmundaceous genera and subgenera are relatively uniformly agreed upon. Phylogenetic relationships among the living osmundaceous genera and subgenera have been less clearly understood (16, 17; 14).

17) was the first to employ phylogenetic systematics methods to resolve relationships among species of Osmundaceae. Using Wagner's Ground Plan/Divergence methodology (34) Miller developed a set of 18 systematically informative morphological characters with at least two states each (17). Results of that analysis more or less supported the accepted segregation of living species into three genera (Osmunda, Leptopteris, and Todea), and resolved the three recognized subgenera of Osmunda s.l. However, T. barbara nested within a clade of Osmunda species, rendering the genus Osmunda a nonmonophyletic assemblage (Fig. 10). In that tree (Fig. 10), species of Leptopteris were least specialized and formed a clade that was the sister group to the remaining Osmundaceous species (Fig. 8 of 17). Osmunda cinnamomea (subgenus Osmundastrum) plus T. barbara formed a small clade attached to the stem at the next more distal node, with remaining species of Osmunda resolving as two clades at the apex of the tree (Fig. 10). The latter included species of subgenus Osmunda (i.e., O. claytoniana, O. regalis, O. japonica, and O. lancea Thunb.) as the sister group to species of the subgenus Plenasium [i.e., O. vachellii, O. javanica l., O. bromeliaefolia (C. Prest.) Kuhn, and O. banksiaefolia (C. Presl) Kuhn].

More recently, 35) assessed relationships among living osmundaceous species using nucleotide sequences of the chloroplast gene rbcL. Included in the analyses were representatives of each of the widely recognized osmundaceous genera and subgenera (i.e., Leptopteris hymenophylloides (A. Rich.) C. Presl., L. wilkesiana (Brack.) H. Christ, T. barbara, Osmunda japonica, O. lancea, O. regalis, O. claytoniana, O. cinnamomea, O. banksiaefolia, O. javancia, O. vachellii). One species of Ophioglossales, one species of Marattiales, and six species of other filicaleans were included as outgroups (35). Analyses were conducted using both maximum parsimony and neighbor-joining methodologies.

Results of the two analyses conducted by 35) were concordant (Fig. 11) except for relationships among species of the subgenus Plenasium. As with the results of the Miller analysis (1971), Osmunda s.l. forms a nonmonophyletic assemblage (i.e., Fig. 1 of 35), but the composition of the clade attached at the basal nodes of the tree differs from the results of 17 (Fig. 11). Whereas a Leptopteris spp. clade is attached at the basal node of the 17 tree and Osmunda cinnamomea + Todea barbara form a small clade attached to the stem at the next node, the 35 results resolve Osmunda cinnamomea as sister to all other species of the family (Fig. 11). At the next more distal node of the Yatabe et al. tree, species of Leptopteris and Todea form a small clade that is the sister group to all of the remaining species of Osmunda s.l. Those results support Miller's recognition of O. cinnamomea as the only species of Osmundastrum, but contradict a basal position for Leptopteris within the osmundaceous clade (Fig. 11).

Our analysis used both the nucleotide sequence characters of 35 and 41 morphological characters, most of which could be scored for all of the taxa (Appendix 1). The results of this analysis (Fig. 12) are fully concordant with those of 35 (Fig. 11) except that T. tidwellii is added to the tree as the sister to T. barbara (Fig. 12). These results strongly support both the appropriateness of assigning T. tidwellii to the genus Todea and the accuracy of the Yatabe et al. tree. They also support segregating Osmundastrum (i.e., O. cinnamomea) to a separate genus (Table 1), as indicated by its location at the base of the tree that includes genera with living species (Fig. 12).

To test whether the much larger number of nucleotide sequence characters in the matrix were swamping a possible alternative topology generated by the morphological characters alone, the nucleotide sequence characters and the nonosmundaceous leptosporangiate taxa were all removed, and the morphological characters were analyzed separately. The results of that analysis yielded four most parsimonious trees of 30 steps. In the strict consensus tree of those results, the two species of Todea resolved as sisters, but the remaining osmundaceous taxa formed a polytomy with the Todea clade. Therefore, while results of the morphological analysis support a sistergroup relationship between T. barbara and T. tidwellii, they do not resolve relationships among the other osmundaceous ferns.

The occurrence of sporangia on laminar pinnules that are similar to those of vegetative fronds in Todea and Leptopteris contrasts with fertile fronds (or frond segments) of Osmunda, which have narrow laminae that are highly reduced as compared to vegetative pinnules (2; 11). This character has added to the perception that Todea and Leptopteris may be basal among living genera of Osmundaceae (e.g., 17) because of their wide fertile pinnules. However, attachment of the Todea + Leptopteris clade at a node above Osmundastrum on the 35) tree (Fig. 12) indicates that the common ancestor of living Osmundaceae could have borne sporangia on pinnules with highly reduced laminae. If the topology of the 35 tree is accurate, then the laminar fertile pinnules of Todea and Leptopteris either represent a derived character among osmundaceous genera with living species or else narrow fertile pinnules evolved independently in O. cinnamomea and in the remaining species of Osmunda (Fig. 11).

Paleobiogeography of Osmundaceae

As early as the Late Permian, unequivocal osmundaceous species (sensu 19) were present in Australian Gondwana (8), with many species occurring in Australia, South America, Antarctica, Russia, and China by the Early Triassic (5; 24). This distribution suggests that the family originated on the Gondwana continent of the Southern Hemisphere during the Permian (8), spreading to the Northern Hemisphere through Asia and South America, and arriving in Europe and North America by the late Triassic (24). The origin of crown-group Osmundaceae by Late Triassic time (19) suggests that some extant genera and species could be extremely ancient. If true, that would help explain the widely disjunct distributions of many living species (e.g., 11; 35). It also would account for the occurrence of Todea tidwellii in western North America, whereas restricted ranges of the two living species are half way around the world in New Zealand and New Guinea.

Calibration of clade ages for Osmundaceae

Osmundaceae and closely related ferns are represented by an unusually large number of morphospecies that range from the Late Permian, through the Mesozoic and Tertiary, and up to the present (8; 17; 32; 23; 29; 24). More ancient dates for the clade (e.g., 22) remain unverified, either because they are calculated from inaccurate tree topologies (21) and/or because the fossils have equivocal systematic relationships to modern Osmundaceae (i.e., Thamnopteridaceae sensu 17).

Fossils of the Osmundaceae and the closely related family Guaireaceae Herbst form one of the most extensive paleontological records (32) for inferring evolution and phylogeny to be found anywhere among fern clades of the euphyllophytes (20). However, inferring the antiquity of crown group Osmundaceae from the paleontological record is complicated by almost all of the fossils being described as morphotaxa, rather than being understood as extinct species of plants. Up through the Jurassic, most morphospecies of compressed fertile osmundaceous fronds do not show synapomorphies for living genera (e.g., Todites Seward; 9). Therefore, without knowledge about the rest of the sporophyte, those species cannot be assigned to a living genus with confidence. Likewise, most anatomically preserved osmundaceous trunks (32; 3; 29) lack diagnostic features of living species, and little is known of their frond architecture, pinnule morphology, sporangia, or spores. Therefore, such fossils could represent either species of extinct genera or extinct species of genera with living representatives.

Compressed fertile frond segments described from the Late Triassic of Antarctica as Osmunda claytoniites Phipps, T N. Taylor, E. L. Taylor, Cuneo, Boucher & Yao (Phipps et al., 1998) provide the oldest unequivocal evidence for the genus Osmunda s.l., and establish a minimum age for crown group Osmundaceae of around 220 million years. Likewise, Todea tidwellii provides a minimum age of ca. 136 million years for the Todea clade. The earliest evidence for a living species of Osmundaceae consists of anatomically preserved trunks of Osmundastrum cinnamomea from the Campanian stage of the Upper Cretaceous (ca. 75 Ma [million years ago]; 15) in western Canada (23). That age establishes O. cinnamomea as having the greatest known species longevity of all vascular plants (23).

Together, the minimum clade ages established by fossil Osmundaceae confirm hypotheses that the crown group evolved by the early Mesozoic, that the Todea+Leptopteris clade was present by the Early Cretaceous, and that living species began to appear no later than the Late Cretaceous. However, if the tree topology presented in Fig. 12 is accurate, then at least some of the Triassic morphospecies of Osmundaceae may actually represent the Osmundastrum clade. In this regard, we eagerly await the testing of these hypotheses by the future reconstruction of osmundaceous fossils as species of extinct plants and by the accurate placement of more fossil species on the phylogenetic tree.