Is the anthophyte hypothesis alive and well? New evidence from the reproductive structures of Bennettitales^†

Secondary xylem tracheary pitting

One of the frequentlycited characters separating cycads from Bennettitales is the common presence of scalariform secondary wall thickenings on tracheids in the secondary xylem of the Bennettitales in contrast to the general view that cycads have circular bordered pits in their secondary xylem (e.g., 28; Table 2). To put this character in the perspective of the anthophytes, “scalariform tracheids in secondary xylem” is not considered to be typical of all the so-called anthophytes. Gnetales typically have circular bordered pits (Table 1), while scalariform thickenings characterize the secondary xylem of many taxa of angiosperms. However, there is more variation in pitting of both angiosperms and gnetophytes than is commonly recognized (64). Although cycads are generally regarded as having secondary xylem tracheids with circular bordered pits, this is not always the case (113). Secondary xylem in several taxa of cycads (e.g., Microcycas A. DC.; 44) has scalariform thickenings and an accurate coding of secondary xylem characters in cycads requires that that condition be scored as such. One can either code cycads as variable for tracheid secondary wall pitting or break the cycads down into groups of taxa (as in 91) that can be coded accurately.

While 108 claimed that bennettitalean wood is characterized by scalariform tracheids, he also noted (110; p. 75) that perhaps “in further” sections pitting might conform with Lignier's observations of Cycadeoidea micromyela Moriere wood in which pitting was “cross-slitted” like “Araucaria and Cordaites newberryi,” That supposition has been verified in more recent studies of bennettitalean stems (e.g., Bucklandia kerae 85).

Cone structure

The so-called bennettitalean “flowers” are compact monosporangiate or bisporangiate cones that consist of a short axis (i.e., a “receptacle”) with determinate growth and that produces helically arranged bracts below the fertile zone (Figs. 1–8). Bisporangiate cones have more-or-less pinnate or variously fleshy microsporophylls attached toward the base of a squat or conical ovulate receptacle. More distally, the receptacle bears erect seeds (i.e., seeds that are terminal on a sporophyll) interspersed among sterile interseminal scales (Figs. 1b, 1d, 1e, 3–5, 7, 8, 17; Appendix 2). The axis has a large pith delimited by a eustele (Fig. 20; 1; 89) from which traces diverge in either a clearly helical arrangement or else are so closely spaced that it is impossible to determine if the traces are helical or whorled (Fig. 9, near top). Several traces enter each bract and microsporophyll. In the ovulate receptacle, eustelar bundles sometimes appear to be reflexed. In such cases, traces to seed stalks and interseminal scales emanate from the reflexed portions of these vascular bundles. Traces to seed stalks and interseminal scales from the stele of the ovulate receptacle typically branch and fuse in the cortex (Figs. 9–11), with a single terete trace entering each ovule-bearing sporophyll and interseminal scale (Figs. 9, 23; 89). We wish to stress that traces to seed stalks (i.e., sporophylls) and interseminal scales are virtually identical, such that the appendage vascularized by any given trace cannot be determined until it actually enters an appendage. Traces that diverge from sympodia of the receptacular eustele (Fig. 20) are collateral bundles that rotate within the ground tissue of the receptacle such that the orientation of xylem and phloem vary from bundle to bundle (fig. 21).

All of the appendages (i.e., enclosing bracts, microsporophylls, interseminal scales and seed-bearing stalks) diverge directly from the axis or ovulate receptacle (Figs. 10, 17), none arising from another or being located in the axil of another (Figs. 18, 19). Although the seeds are nearly completely surrounded by and variously enclosed by interseminal scales (Figs. 7, 8, 12–19), extremely well-preserved cones (e.g., specimens in Figs. 6, 8, 18) verify that there is no evidence of a specialized “cupule” or other additional enclosing structures (Figs. 15–19). This last point is in agreement with the recent interpretations of 80, 93, and 37 and is addressed more completely in the discussion and phylogenetics sections of this paper.

All these characters are consistent with the interpretation that bennettitalean reproductive structures are simple cones wherein both microsporangiate and megasporangiate structures are foliar in nature (i.e., sporophylls). Bracts that enclose the fertile organs (Figs. 1b, 1c, 2f, 5) are linear and display several vascular bundles above the base. Interseminal scales are radial in construction, columnar in shape, and polygonal in cross section (Figs. 18, 19). They display a terete xylem strand (Figs. 15; 16, at arrowheads) that shows no evidence of a maturation pattern. In some interseminal scales, the vascular bundle divides in two (Fig. 15, upper left arrowheads). The xylem is surrounded by a combination of parenchymatous and sclerenchymatous ground tissue (Figs. 15, 16, 18, 19) that sometimes displays resin canals. It is not known whether the interseminal scale bundle is collateral or radial. Except when fused to an adjacent structure, there is a distinct cutinized epidermis at the periphery of each interseminal scale (Figs. 19, 26).

Microsporophylls/synangial/sporangial position

There are a number of undoubted bennettitalean microsporophylls known from direct attachment to reproductive branches, as components of bisporangiate cones, or that have clear synapomorphies of Bennettitales (110, 104; 23, 24). These vary in structure and in themselves raise some interesting questions about sporangial position, microsporophyll homologies, etc. (figs. 1, 2). There are also a number of compressed morphotaxa that have been assigned to Bennettitales (e.g., Leguminanthus 53) or considered as immediately relevant to their systematic position (e.g., Piroconites), that have sometimes been included in phylogenetic analyses of Bennettitales (e.g., 14), but their affinities are somewhat problematical. They often are relatively poorly preserved, difficult to interpret and several do not appear to have synapomorphies that allow an unequivocal assignment to Bennettitales (e.g., Leguminanthus). These fossils have complicated the understanding of bennettitalean microsporophylls when encoded for phylogenetic analysis in a way that introduces character states presumed present because of assumed affinities with Bennettitales. Some of these morphotaxa comprise genera with remarkably divergent characters. Such fossils are discussed more fully in Appendix 2.

Ovules and seeds

Bennettitalean seeds and their associated structures have been the object of considerable discussion since the time of 56, especially in the context of possible links between Bennettitales and Gnetales. Among Bennettitales, the seeds of Cycadeoidea, Williamsonia, and Cycadeoidella vary from one another in several details, but anatomically preserved specimens of excellent preservation reveal that all have a similar basic construction. Seeds are erect and terminate a narrow sporophyll (Figs. 1, 2, 5, 7, 8, 10, 17, 23) that consists of a terete xylem strand that is surrounded by several layers of thin-walled cells (Figs. 2f, 6). In cross sections, the seeds are radial, being more-or-less circular in Williamsonia (Figs. 16, 18, 24–26; summarized by 93) and Cycadeoidella (67), and ranging from angular to stellate (Fig. 15) at different levels in Cycadeoidea (=Bennettites Carruthers of some workers; summarized by 80). The seeds of most species taper to a narrow, round micropylar opening (e.g., Fig. 24), which depending on the species, may or may not protrude above the tips of the surrounding interseminal scales (Figs. 2f, 8, 10, 12–14, 17, 23). In contrast to some gnetalean seeds (e.g., 5), excellently preserved permineralized bennettitalean seeds show no evidence of cellular proliferations or secondary cell divisions that occlude the micropyles of mature specimens.

There is a single multilayered integument that is unvascularized and that is differentiated into several types of cells (Figs. 12–16, 22–24). The inner epidermis of the integument typically consists of distinctive radially elongated cells near the tip of the micropylar canal (Fig. 1g, 2g, 2i, 22–24), but cells of that layer are isodiametric at more proximal levels (Figs. 22, 23, 25). The nucellus separates from the integument at the base of the seed (Figs. 1g, 2g, 2h; 13, at arrows; 14, at arrow) and is vascularized by a cup-shaped bundle of tracheids that terminates the vascular bundle of the stalk. Distally, the apex of the nucellus forms a cylindrical- or conical-shaped parenchymatous plug (Figs. 1g, 2i, 22, 23, 25). In both immature and full-sized seeds, the nucellar plug is typically tightly inserted into the base of the micropylar canal (Figs. 12–14, 22, 23). In contrast to Gnetales, and nearly all other gymnospermous seed plants, there is no pollen chamber. Within the nucellus, many bennettitalean specimens preserve both tissue of the megagametophyte (e.g., fig. 4D of 93) and an embryo that is dicotyledonous at maturity (Figs. 12, 13). As is also the case for Gnetales and many flowering plants, the megaspore membrane is either extremely thin or not preserved at all in bennettitalean seeds (e.g., Fig. 26).

This relatively simple, but distinctive seed structure is clearly recognized from well preserved permineralized seeds (e.g., Figs. 13, 14), and from immature ovules in which specialized cells of the integument have not yet differentiated (Figs. 17–26) and the nucellus fills the seed cavity below the micropyle (Figs. 1g, 22). In some ovules, meiosis has yet to occur or forms a linear tetrad (Fig. 22, see the possible example at arrowhead; 17), whereas in others, there is an area at the center of the nucellus that represents the megagametophyte at a stage before cellularization has occurred (Fig. 26, at m).

Ovule enclosing structures

If the Bennettitales is a closely related or sister group to flowering plants, then it is reasonable to expect that bennettitalean cones and seeds may provide evidence for the origin of the double integument and/or carpel or for structures that may have given rise to a second integument and carpel. Early workers (56; 110; 95) recognized that there is a cup-shaped covering of the seed stalk and integument of Cycadeoidea seeds, but disagreed about the nature and significance of that structure. By comparing that structure to the cupule of Lagenostoma, 56 set the stage for some future workers to interpret Bennettitales as cupulate plants (e.g., 45; 12, 13; 31). However, other authors have interpreted the same structures differently. 95 provided evidence that the outer covering was originally in organic connection to (or a part of) the seed integument, prompting 80 to interpret it as a sarcotesta of tubular cells. It is also possible that these tubular cells of the sarcotesta in Cycadeoidea are a dense indument of large trichomes, an interpretation that is consistent with anatomical evidence (e.g., Fig. lf).

67 recognized a thin, cutinized membrane on both seeds and interseminal scales of Cycadeoidella japonica (i.e., figs. 14–17 of 67) and interpreted that structure as a cupule. While that interpretation is plausible, cupules of a similar structure are not known from other bennettitalean fructifications. An alternative interpretation is that the putative cupule represents the cuticle of seeds and interseminal scales in which cells beneath the cuticle are incompletely preserved.

A third category of structure that has been interpreted as a “cupule” in bennettitalean cones is formed by interseminal scales. Many authors have recognized that interseminal scales of bennettitalean fructifications form a ring around each seed (e.g., figs. 16, 17 of 80; fig. 2A of 93). In most examples, the ring consists of what are clearly adjacent interseminal scales (Fig. 19), but in some compressed specimens the surrounding structure appears to be a single entity (i.e., Bennetticarpus crossospermus Harris; 12). In Bennetticarpus Harris, the putative cupule is quite thick, possibly reflecting its origin from several interseminal scales.

Anatomical evidence in support of such an hypothesis possibly could be provided by the cone of Williamsonia bockii (93). In W. bockii, adjacent interseminal scales are not separated by a pair of cuticles. Rather, the peripheral cells of adjacent interseminal scales interdigitate, forming the characteristic anatomical fingerprint for postgenital fusion. Although postgenital fusion is a developmental phenomenon widely considered to be a hallmark of angiosperms (e.g., Endress 2001), sections of W. bockii reveal that it also occurs in the Bennettitales (93). While this co-occurrence could be inferred as providing evidence for bennettitalean/flowering plant relationships, the possible distribution of this character in other fossil taxa has not been investigated. Also, there currently is no evidence that the ontogenetic pathway leading to fused interseminal scales in W. bockii is the same as for postgenital fusion of angiosperm carpels. Thus, putting too much emphasis on this one character might be premature. In addition, unlike the hypothesized structure of Bennetticarpus, all the interseminal scales of W. bockii are fused to each other, not just those that surround seeds (fig. 2A, 2B of 93). Despite forming a continuous tissue, in surface views of W. bockii the individual interseminal scales and seed micropyles are clearly evident as distinct entities (fig. 8 of 93).

A much thinner cupule has been interpreted as occurring in Vardekloeftia sulcata Harris (45; 12; 75), but a more recent interpretation accounts for all of the cuticular layers of Vardekloeftia as being derived from the integument and nucellus (80). More recently still, 37 also have reinterpreted Vardekloeftia as lacking a cupule. At the present time, there is no substantive evidence from bennettitalean fossils for the production of a seed-enclosing structure that is comparable to any of the heterogeneous structures called “cupules” in hydrasperman seed ferns, Corystospermales, Petriellales, or Caytoniales (79; 49; 37; 94, pp. 323–335 in this issue).

Analysis 1

We reanalyzed the matrix of 37 to confirm their results using an alternative program (i.e., NONA, rather than PAUP). Our analysis yielded 1180 most parsimonious trees of 314 steps (CI = 45, RI = 80) with a majority rule consensus topology (not figured) that is roughly equivalent to the 55% majority rule consensus tree figured by Friis et al. (i.e., fig. S2 of 37). As in the results of the 37 analysis (strict consensus tree not figured by 37), the strict consensus tree of our results has a large number of collapsed branches, thus resolving few clades (Fig. 28) and confirming that the general tree topology obtained by 37 is also obtained using the NONA program. The tree is rooted by the homosporous progymnosperm Tetraxylopteris Beck, with successive nodes along the stem including (1) a polytomy of heterosporous progymnosperms; (2) a huge polytomy of seed ferns, cycads, cordaites and conifers, Pentoxylales, Bennettitales, taxa of the Gnetales, Erdtmanithecales; and (3) Caytonia + flowering plants (Fig. 28).

More specifically, in the parsimony-based strict consensus tree based on our analysis, relationships among Ephedra, Welwitschia, Gnetum, the charcoalified seeds (of 37), the Erdtmanithecales, and the Bennettitales are all unresolved (arrows in Fig. 28), thereby agreeing with the strict consensus of the most parsimonious trees obtained by 37 that a clade including Gnetales and Bennettitales is not resolved. Only when parsimony is relaxed by the use of majority-rule consensus is the figured resolution of a clade that includes Bennettitales nested among gnetophytes (i.e., fig. S2 of 37) approached. Furthermore and as discussed below, it appears that the majority rule consensus may have been misapplied (or over-interpreted) in the 37 analysis (see explanation below).

Analysis 2

Next, we reanalyzed the matrix of 49 with the charcoalified fossil seeds added and coded as in 37, but with the “Erdtmanithecales” omitted. This allowed us to test what effects the addition of the charcoalified seeds have on the inferred relationships among Bennettitales and Gnetales without the added variable of an additional new taxon (i.e., Erdtmanithecales).

Our analysis yielded 357 most parsimonious trees of 313 steps (CI = 45, RI = 80) and produced a strict consensus tree that has 23 collapsed branches and better resolution than in the results of analysis 1 (Fig. 29). The topology of this tree is less fully resolved but otherwise roughly equivalent to the results obtained by 49. Major groups of clades are attached to the stem of the tree in roughly the same order that is found in previous analyses by most authors (Fig. 28; 68; 79; 28; 49), and the charcoalified seeds of 37 form a polytomy with the extant gnetophytes (Fig. 29, at upper arrow).

The tree is rooted by the homosporous progymnosperm Tetraxylopteris, with the polytomies at successive nodes along the stem including (1) heterosporous progymnosperms, (2) a large polytomy containing hydrasperman seed ferns + medullosan seed ferns + higher seed ferns + cycads + cordaites and conifers, and (3) a smaller polytomy containing glossopterids + Pentoxylon + gnetophytes (including the charcoalified seeds of 37) + a clade consisting of Bennettitales + (Caytonia + flowering plants) (Fig. 29). These results contradict the conclusion of 37 that the addition of the charcoalified seeds to the matrix of 49 produces a topology in which Gnetales resolve as a monophyletic group that includes Bennettitales (Fig. 29). In contrast, our results are consistent with the distinct differences between the cones and seeds of Bennettitales and Gnetales that are described earlier in this paper, as is indicated by the attachment of the Gnetales (plus charcoalified seeds) clade and the Bennettitales clade in separate clades (arrows in Fig. 29). These results suggest that the topology figured by the 55% majority-rule consensus tree of 37 may derive from addition of the taxon “Erdtmanithecales.”

Analysis 3

To test whether the results of our analyses (and those of 37) are influenced by the construction of the character matrix being tested, we combined the matrices of 79 and 49 and removed all unambiguously redundant (overlapping) characters, leaving one in each case. Character 63 was modified from 79; character 37 “micropyle” modified to “tubular micropyle” with three states: [0] missing, [1] composed of two integumentary layers [2] composed of three integumentary layers). The taxa included in this matrix consist of a compromise between those included in 79 and 49, with the goal of maintaining as high a character state/taxon ratio as possible in the matrix.

Laminar venation character states (character 8) were modified following 68 to more accurately reflect the morphologies of the taxa demonstrating the divergent venation conditions. Instead of one alternative “reticulate,” there are two: “reticulate hierarchical” and “reticulate anastomosing,” the latter reflecting anastomosing venation of similar order veins. Caytonia was coded as “reticulate anastomosing” as was “Glossopteris.” Zamiaceae was coded as multistate for character 10 reflecting actual variation in guard cell configurations (e.g., 102). Cataphylls (character 7) were coded as absent for Caytonia. Also, character codings for Caytonia were modified to remove embedded hypotheses in character codings as discussed in 68 and 79. Codings for the characters of Caytonia are (1-12-1-1011?0??-210—-10--0——000001100-00-110-02202-0100-10111--1—2106——-111100000—-1————).

Results of this analysis yielded 12 most parsimonious trees of 328 steps (CI = 50, RI = 77), and only five branches collapsed in the strict consensus tree (Fig. 30). This highly resolved tree places Caytonia (at arrowhead) among other seed ferns and cycads at a polytomy near the base of the tree. Ginkgo occurs within a paraphyletic assemblage of cordaites and conifers at four adjacent nodes on the stem of the tree, but otherwise the tree was free of novel topologies. As in the results of analyses 2 and 4, the charcoalified seeds nest within the gnetalean clade, and Bennettitales is attached to the stem at a separate node (arrows in Fig. 30).

Analysis 4

For the next analysis, two morphotaxa hypothesized by 38 and 37 to be the seed- and pollen-producing organs belonging to the same order of plants (i.e., Erdtmanithecales; 38) were added to the matrix for analysis 4 and scored as in 37, but without pollen characters for Erdtmanispermum because, as discussed below, pollen in micropyles does not necessarily belong to the same taxon as the seeds with which they are associated. This analysis was designed to test both the putative assignment of these two morphogenera to the same group of plants and also to assess the impact of their inclusion on the cladistic relationships of Bennettitales and Gnetales.

Results of analysis 4 yielded 151 most parsimonious trees of 330 steps (CI = 50, RI = 77), and 12 internodes collapsed in the strict consensus tree (Fig. 31). Relationships among the taxa were relatively similar to, but somewhat more highly resolved than in the results of analysis 1 (Fig. 28). Notable differences from the results of analysis 1 were the placement of Caytonia within the large polytomy that included “higher” seed ferns, Bennettitales, Glossopteris, cycads, cordaites and conifers, and Gnetales (Fig. 31). The two taxa of the hypothesized Erdtmanithecales did not nest together in the strict consensus tree of our results. Whereas Erdtmanispermum nested within the gnetophyte clade, Eucommiitheca was not part of that clade (Fig. 31, at arrows), but rather part of a small polytomy at an adjacent node on the stem. The latter polytomy included Bennettitales, Pentoxylon, Eucommiitheca, and flowering plants (Fig. 31).

Erdtmanispermum and Eucommiitheca were never sister groups in any of the most parsimonious trees unless pollen characters are added to the matrix for Erdtmanispermum. Thus, following the methodology explained by 40; i.e., comparing the results of analyses that treat the different organs as a single taxon with results from analysis that list them as separate taxa) and in the absence of additional information, these data do not support the hypothesis that Erdtmanispermum and Eucommiitheca are morphotaxa of the same clade.

While the methodology of phylogenetic context (40) used to position Erdmanispermum and Eucommiitheca in analysis 4 (Fig. 31) may, in speculation, have weaknesses in cases where the combination of organs of a putative fossil taxon represents extreme variation from character complements represented in all known taxa (mosaics), there have been no compelling tests of this concept with fossil taxa. Such speculation depends on the assumption that these separate organs will be defined, principally, by the characters that separate them and that other characters (possible synapomorphies) that may be present in these fossils will fail to nest these morphotaxa even in the context of a more complete matrix. Thus while untested, it is possible that separate organs of the same taxon would not nest together in phylogenetic context under certain circumstances, yet, in the absence of other compelling evidence (see discussion below), there is really no other objective test for the hypothesis that these disparate organs represent the same taxon. Of course, in cases where the relationship between those organs remains in question, as with Erdmanispermum and Eucommiitheca, this “phylogenetic context” approach both tests the combined organ taxon concept and identifies the most probable alternative relationship for each organ.

In the results of this analysis, Erdmanispermum nests with the charcoalified seeds of 37 and the living genera of Gnetales, with which it shares numerous characters. Eucommiitheca, on the other hand, forms a polytomy with Bennettitales, Pentoxylon, and flowering plants, which reflects a set of characters that do not conform more closely to any taxon in the polytomy than it does to the others.

Analysis 5

The matrix for analysis 5 was the same as for analysis 4, except that the character codings for Erdtmanispermum and Eucommiitheca (including pollen characters) were included within a single taxon “Erdtmanithecales.” Results of analysis 5 yielded 252 most parsimonious trees of 334 steps (CI = 49, RI = 77), and 19 internodes collapsed in the strict consensus tree (Fig. 32). Relationships among the seed plant taxa were far less completely resolved in this strict consensus tree than in the strict consensus tree of results from analysis 4 (Fig. 31). The monophyletic group reported by 37 that included Gnetales + Bennettitales + Erdtmanithecales and “Charcoalified Seeds” (i.e., the “BEG” clade of 37) did not resolve in the strict consensus of the most parsimonious trees (Fig. 32).

The discrepancy between results obtained here and those reported by 37 is due in part to over-interpretation of a majority rule consensus tree of phylogenetic results by 37, see their Supplementary Information). They used this consensus method to support the existence of a BEG clade (Bennettitales + Erdtmanithecales + Gnetales + charcoalified seeds; see their figs. 3 and S2). However, the BEG clade was not seen in all most parsimonious trees (41.9% of MPTs; see their Table II), a conclusion supported here by our reanalysis of their data (analysis 1). While the majority rule consensus method is useful, for example, for reporting branch support values from jackknife or bootstrap analysis, it is invalid to present a majority-rule consensus of the MPTs and report it as if it supports either a particular phylogeny or a particular clade (see e.g., 87). Some of their MPTs, which are equally valid interpretations of the data, conflict with the arrangement that they prefer (i.e., the BEG clade). However, in a reanalysis of their data that focused only on characters not missing data for Erdtmanithecales (about 25% of the total), the BEG clade was seen in nearly all MPTs (∼99%; their Table II) and also had weak bootstrap support there (60%).

Among most parsimonious trees from analysis 5 were those with Bennettitales and Erdtmanithecales nested within a clade that also included the living gnetophytes and the charcoalified seeds of 37 (i.e., the BEG clade, 37). These, however, were the minority of trees numerically (18 trees had such a monophyletic group) and trees with a clade consisting of Gnetales+charcoalified seeds+Erdtmanithecales (i.e., with Bennettitales elsewhere) were more common, as were trees with a clade consisting of Gnetales+Erdtmanithecales (i.e., with Bennettitales and charcoalified seeds elsewhere). While some of the most parsimonious trees (those with the resolved BEG clade, could possibly be interpreted as supporting the hypothesis that the reproductive structures of Bennettitales are comparable to those of Gnetales; 37), the structural differences between Bennettitales and Gnetales detailed earlier in this paper suggest than an alternative explanation may be more consistent with the loss of resolution of this clade in the strict consensus tree of analysis 5. The appearance of the BEG clade even in relatively few most parsimonious trees (7%) makes analysis 5 unique among the results of our other tests. However, the occasional appearance of a BEG clade is not correlated with the addition of the charcoalified fossil seeds to the matrix, but rather with the addition of Erdtmanithecales, a taxon of contestable validity.

“Charcoalified seeds”

All the results except those from analyses 1 and 5 (where a gnetophyte clade was not resolved) yielded strict consensus trees that included “charcoalified seeds” in a monophyletic group with Gnetales (Figs. 29–313031). Likewise, all of the results except those from analyses 1 and 5 (which show little resolution in much of the tree) resolve Bennettitales and Gnetales as discrete clades that are part of different larger clades (Figs. 29–313031). Our results support the interpretation that the charcoalified seeds are additional representatives of Gnetales. However, because the charcoalified seeds uniformly nest within the gnetalean clade (when a gnetalean clade is resolved), they neither provide additional support for a morphologically defined “anthophyte”clade, nor do they help resolve relationships among Gnetales and Bennettitales.

Gnetales vs. Bennettitales

This review, including the analyses described above, detailed descriptive data, and the extensive discussion of comparative seed morphology, suggests that Bennettitales are taxonomically distinct from Gnetales and that there are no unequivocally interpreted fossils to link the groups in a way that would strengthen the “anthophyte” concept in phylogenetic analyses. Moreover, with respect to remaining gnetalean features, there are numerous differences other than contested seed characters (Table 1) and these have been recognized for a long time. For example, Thompson noted in 1912 (p. 1099): “It may be stated at once that on the anatomical side there is very little evidence for connecting the Bennettitales with Ephedra, although this genus, being the most primitive of the Gnetales, is the one where the evidence ought to be found…almost every tissue presents grave obstacles to this view.”

Erdtmanithecales

37 coded Erdtmanithecales as a composite of the characters found in at least five separate fossil taxa that occur in different localities from beds of different ages spanning the Early to Late Cretaceous interval (74; 38). However, we wish to caution that associating dispersed organs to create a composite taxon poses a set of challenges (most famously perhaps exemplified by the farrago taxon Brontosaurus [vs. Apatosaurus], and in paleobotany, by early reconstructions of various Devonian plants, most dramatically the association of different organs in 51 reconstruction of Asteroxylon). There is always a risk of error in assembling taxa from disparate organs even if, as in this case (37), certain plant organs are linked by somewhat similar pollen grains. This is because the actual taxonomic distribution of such grains may not be fully appreciated based on available fossil evidence, and unrelated or distantly related taxa may have been lumped together into a single chimerical entity.

Seeds and pollen producing structures upon which the concept of Erdtmanithecales is based are associated by the occurrence of similar pollen in both. However, there is evidence from both living and fossil plants (e.g., 46) that documents that this pollen-based association alone may be misleading. Of special concern in this case is the fact that both wind- and insect-dispersed pollen tend to collect on the stigmas or micropyles of other species, sometimes to the extent of interfering with their reproductive biology (e.g., 83; 66).

74 discuss this issue and stress the consistency with which the same kind of pollen is found exclusively within the micropyles of the Danish specimens of Erdtmanispermum. However, there are important differences between these pollen grains and those found in the pollen sacs of Erdtmanitheca texensis. Those differences leave doubt as to what the pollen-producing organs associated with Erdtmanispermum might actually have been like and therefore as to the nature of the taxon that produced them. Such doubt is strengthened further by the distinct possibility that Eucommiidites-type pollen may have multiple origins (97). Inherent reservations concerning the practice of composing taxa based on pollen association rather than on organic connection or on a broader common suite of characters are heightened in the case of the Erdtmanithecales because the dispersed organs occur at different locations (e.g., in Texas vs. in Denmark, 74). Specimens assigned to Erdtmanithecales are also of disparate ages, Early Cretaceous vs. Late Cretaceous, in the subset of taxa identified by 74 as representing Erdtmanithecales. Moreover, there are notable differences among erdtmanithecalean taxa that even include contrasting morphologies e.g., Eucommiitheca vs. Erdtmanitheca vs. Bayeritheca Kvaček & Pacltová (fig. 6 in 38) and phyllotaxy (54; 37).

Finally, there appear to be notable differences between the Eucommiidites pollen grains found in the pollen-bearing organs as compared to those found in the micropyles of the seeds assumed to belong to the same taxon, but those differences are not reflected in the character selection or codings of 37. Pollen grains from Erdtmanitheca texensis 74 have smooth tectal ornamentation with very fine tectal depressions that do not appear to penetrate the tectum (figs. 2A–2F, 3G in 74). In contrast, pollen grains found in the micropyles of the seed taxon Erdtmanispermum are arguably foveolate (fig. 5G in 74) with tectal perforations (fig. 6 in 74). In addition, pollen grains in the seed micropyles are dramatically different from grains of Erdtmanitheca with respect to certain ultrastructural characters that have been considered taxonomically significant (e.g., 105). Pollen from Erdtmanitheca has a poorly defined footlayer, a relatively thin granular layer and an extraordinarily thick tectum (fig. 3D–3G in 74). In contrast, pollen found in the micropyles of Erdtmanispermum, in addition to having tectal perforations (fig. 5G in 74), has a remarkably thick footlayer, a granular layer with larger granules than those of Erdtmanitheca, and a relatively thin tectum (fig. 6 in 74).

A conservative approach to testing relationships among disparate fossil morphotaxa is to enter the unequivocal characters of individual fossil organs that putatively represent a single taxon into a phylogenetic analysis to gain insights into how the organs are related to other seed plants and to each other (40). And as discussed more briefly above, in analysis 4 we conducted such a test of the Erdtmanithecales. That analysis included two representative taxa of putative Erdtmanithecales (i.e., Erdtmanispermum Pederson & Friis without pollen characters, and Eucommiitheca Friis and Pederson with pollen characters) in the combined Hilton and Bateman/Rothwell and Serbet analysis. Results of that analysis do not support the existence of the Erdtmanithecales. The strict consensus tree of those results nests Erdtmanispermum with a monophyletic group that also includes Gnetales and the “charcoalified seeds,” while Eucommiitheca is part of a polytomy with Bennettitales and Pentoxylon that attaches to the stem of the tree at a different node (Fig. 31, at arrows).

Thus, although the strict consensus tree of our results for analyses 3 (Fig. 30) and 4 (Fig. 31) did resolve an anthophyte clade, there were no circumstances among the results of our various experiments (Figs. 29–313031) under which including the “charcoalified seeds” and/or the taxon Erdtmanithecales added even minimal support for that topology. (Figs. 28–3229303132). Unless and until future studies verify the validity of the hypothesized Erdtmanithecales, we need to exercise caution in intepreting the results of analyses that include that taxon. Under such circumstances (i.e., no support for the BEG clade in strict consensus trees), inclusion of Bennettitales within a gnetophyte clade remains tenuous at best.

Caytoniales and the origin of angiospermy

The order Caytoniales 101 is based on the concept of a Caytonia plant, which is derived from an assemblage of compressed organs with excellent cuticular preservation that are associated with each other at the same localities and linked by a number of anatomical characters (101; 46, 47). Associated organs include dispersed leaves, pinnate pollen organ-bearing structures, and megasporophyll-like branching systems terminating in ovulate cupules (101; 46, 47; 12; 92; 96). Larger parts of the plant with interconnected organs and/or internal anatomy have not been found, and on the whole, there are relatively few characters defining the taxon (i.e., available for cladistic analysis).

Caytonia has been associated with angiosperms through various hypotheses since the time of 101, who interpreted the cupules to be angiosperm carpels. Carpels of course, along with a second integument are key angiosperm-defining morphological characters and would be compelling evidence of angiospermy if convincingly demonstrated in any fossil. However, Caytonia has been removed to the seed ferns following the discovery by 46 of pollen grains within micropyles of the cupulate seeds (i.e., the absence of angiospermous pollination). Nonetheless, the general similarity between the incurved cupules of Caytonia and the anatropous angiosperm ovule has inspired a number of successive hypotheses purported to link angiosperms and Caytoniales (41; 28, 29, 30). Building on this similarity, Doyle has hypothesized a multistep model to explain the origin of the bitegmic ovule plus carpel combination from a Caytonia-like ancestor. That model (30) involves reduction of ovule numbers within the caytonialean cupule and the de novo origin of the enclosing carpel from the rachis of the caytonialean “megasporophyll”.

Doyle's hypothesis is well reasoned and an appealing synthesis of structural, evolutionary and developmental data. As engaging as it may be, however, that hypothesis relies on more basic hypotheses about Caytonia structures (e.g., abaxial or adaxial identity of the inner cupule surface; 30) that have yet to be verified by anatomical or developmental evidence. In addition, that hypothesis does not sufficiently explain the dramatic differences in some characters between Caytonia and angiosperms, especially pollen morphology and ultrastructure (114; 68), both of which have well known systematic significance. There is also significant variation in leaf venation that is minimized by codings that offer only one alternative for net-veined (68). Finally, the hypothesis is not supported by any known fossil intermediates. In the context of these observations, the coding of Caytonia characters for the systematic analyses performed in this study has been done with care to avoid including possible hypotheses of structure. Results of those analyses (Figs. 30, 31) agree with the results of many earlier phylogenetic studies of seed plants (e.g., 12; 68; 79) by removing Caytonia from the clade that includes Bennettitales, Gnetales, and flowering plants, and nesting it among other “seed fern” taxa much lower on the stem of the tree (e.g., Fig. 30). This approach emphasizes the need for the development of more compelling evidence for the structure and homologies of Caytonia fertile structures and provides a conservative estimate for what we actually know (and what we don't know) about the origin of the angiospermous outer integument and carpel.

Cycadales vs. Bennettitales

Relationships among cycads and Bennettitales either are unresolved (Figs. 28, 32), or the two taxa remain separated in each of our analyses. In the most highly resolved trees Cycadaceae + Zamiaceae forms a polytomy with the “higher seed ferns” (i.e., Callistophyton, Peltasperms, Corystosperms, Glossopteris, Caytonia) and Bennettitales are attached to a higher node on the stem of the tree (Figs. 29–313031). While certain key homologies of the reproductive structures of cycads vs. Bennettitales remain unclear (ovulate receptacle for example), available and newly available data (Table 2), continue to support the absence of close systematic relationships between these two clades.

Unconfirmed competing hypotheses for flowering plant origins

There are two or three categories of currently competing hypotheses for the origin of angiosperms that are derived from the study of living and fossil plant form. One, as represented by the anthophyte hypothesis, proposes that the outer seed integument and carpel are derived from fertile structures that already were aggregated into a flower-like reproductive organ (i.e., a strobilus or cone). That hypothesis, asserting that a strobilus or flower-like aggregation of fertile organs was inherited from a common ancestor by angiosperms and its sister groups, is most commonly tested by looking for similarities and transformational series among the fertile organs of angiosperms and hypothesized angiosperm sister groups (e.g., Gnetales, Bennettitales). As elaborated in this paper, that hypothesis has yet to find support from the similar structure of potentially homologous fertile organs among Bennettitales, Gnetales, and flowering plants. Indeed, as elaborated earlier in this paper, the putatively homologous fertile organs of Bennettitales and Gnetales (particularly cone and seed structure) are very dissimilar. Moreover, the paleontological record has yet to yield fossils to help form a transformational series of morphologies between the fertile organs of Bennettitales, Gnetales, and flowering plants.

The second category of hypotheses, as represented by interpretations of the ovulate cupules of Caytonia and elaborated by Doyle and others (e.g., 30), infers that the angiosperm outer seed integument and carpel are derived from modifications of a seed fern megasporophyll, with aggregation of the fertile parts into a flower presumably having occurred later. As exemplified by the seed-bearing cupules of Caytonia, and in the context of associated assumptions, that hypothesis appears to be quite plausible. However, crucial structural features of the Caytonia cupule and cupule-bearing “rachis” upon which that hypothesis relies have yet to be confirmed by fossils of adequate preservational mode and quality. We reiterate that when the characters of Caytonia are coded in a conservative fashion with respect to equivocal structural features (e.g., analysis 3), Caytonia relocates from the sister group to flowering plants (e.g., Fig. 28) to a much lower position on the spermatophyte tree (e.g., Fig. 30). Up to the present, the hypothesized transformations of structure leading from the Caytonia cupule to an angiosperm carpel have not been confirmed by either developmental transitions or transformational series of mature morphologies.

A third set of hypotheses focuses on the origin of the angiospermous carpel and outer seed integument by development of ovules at positions occupied by pollen-producing structures (i.e., microsporangia) in the putative immediate ancestor of flowering plants. Such hypotheses are reminiscent of Iltis’ famous “catastrophic sexual transmutation” theory for the origin of maize (50), and include the “gamoheterotopy theory” of 62, 14) and the “mostly male theory” of 39. The gamoheterotopy theory focuses on transformation of synangiate microsporophylls (Figs. 1b, 1e) to seed-bearing megasporophlls at the apex of the cone of Bennettitales without there being a series of morphological transformations to document the change (63). Therefore, that hypothesis predicts that it is not testable using the fossil record.

The mostly male theory utilizes the distribution of a specific LFY gene paralog in one of the hypothesized topologies of extant seed plant trees to infer that the flowering plant carpel may have originated in the transformation of the apical microsporophylls of indeterminate fructifications to megasporophylls of a fructification with determinate growth (see 39 for a detailed explanation). 39 consider many characters of Caytonia to be too dissimilar to those of angiosperms for Caytonia to be a plausible ancestor, an opinion with which we agree. Those authors concentrate on the Corystospermales as a more plausible angiosperm sistergroup (90). Many of the corystosperm characters emphasized or hypothesized by 39 are poorly known or unverified by evidence from the fossils, and other features (e.g., attachment of seeds to the abaxial surface of the cupule in corystosperms; 52) are difficult to reconcile with transformation of a corystosperm cupule to an angiosperm carpel (see also 94, pp. 323–335).

The mostly male hypothesis also relies on several nested hypotheses that remain untested. Nevertheless, that hypothesis does predict the presence of ectopic ovules on microsporophylls in apical regions of the fructifications that are transisional to ancestral flowering plants, and therefore at least one aspect of the hypothesis is testable using the fossil record. Up to the present time, no fossil fructifications with the “transitional morphologies” predicted by the mostly male hypothesis have been discovered in the fossil record.

Anthophytes, and alternative sets of characters available for morphological or nucleic acid sequence analyses

The reality of an anthophyte clade has been repeatedly challenged based on analyses of DNA sequence data (e.g., 10; 7), and remains as much a question as it did before we conducted the analyses discussed above, and before those analyses that incorporated “charcoalified seeds” (37). Because the “charcoalified seeds” described by 37 are cleanly nested with Gnetales in all analyses to date in which a gnetophyte clade is resolved in the strict consensus tree of most parsimonious trees (i.e., 37; Figs. 29–313031 of this paper), those “charcoalified seeds” do not lend support for the anthophyte hypothesis.

The results of almost all nucleotide sequence based analyses remove Gnetales from the anthophytes (7), suggesting that many of the characters commonly cited as linking the two groups may be convergent. There is a distinct malleability of all morphology based data sets manifest to anyone who has done such analyses. This malleability is based in part on the inherent subjectivity in selecting and scoring characters. On the other hand, taxon sampling for analyses based on nucleic acid sequences is currently restricted to only the small fraction of clades that are not yet extinct (15; 78). This results in domains of available characters that are largely disjunct for morphological and nucleotide sequence based analyses (Fig. 33). These constraints suggest advantages to combining molecular with morphological data sets in analyses aimed at assessing the reality of an anthophyte clade. Such an approach adds a measure of objectivity and broader sampling (save for extinct groups) and allows for the incorporation of valuable information presented by fossil evidence.

Nonetheless, at this time, and based on the data sets we used, our analyses do consistently support an anthophyte topology of the spermatophyte tree (e.g., Fig. 30). This may have been expected because each of the matrices used (or modified) herein to evaluate the significance of bennettitalean characters and the “charcoalified seeds” reported by 37 already supported a robust anthophyte clade. However it is clear that the constituents of that clade are tightly circumscribed and that connections among them remain as cryptic as ever.