Experimental Cladistic Analysis of Anatomically Preserved Arborescent Lycopsids from the Carboniferous of Euramerica: An Essay on Paleobotanical Phylogenetics

Abstract

This evolutionary cladistic analysis of the arborescent (wood-producing) lycopsids, an exclusively fossil group of vascular plants, is confined to the strongest available data: anatomically preserved fossils that have been painstakingly reconstructed into conceptual whole plants. Ten Carboniferous genera are represented by 16 species: four pseudoherbs/"shrubs" and 12 of the arboreous (tree-sized) species that epitomize the Pennsylvanian coal swamps of Euramerica. The 69 vegetative and 46 reproductive characters are described in detail; several key terms are redefined and homologies reassessed. Binary coding was imposed throughout the data matrix, which contained only 5% missing values despite limited X-coding. Lack of an acceptable outgroup necessitated construction of a hypothetical ancestor for character polarization and tree rooting. Our experimental approach analyzed the full data matrix plus four submatrices (growth habit characters excluded, Chaloneria excluded, vegetative characters only, reproductive characters only) and screened topologies of subminimal as well as minimal length. Interpretation focuses on the ten monophyletic genera and marginally favors the topology ((Paurodendron, Oxroadia) (Anabathra (Chaloneria (Sigillaria ((Diaphorodendron, Synchysidendron) (Hizemodendron (Lepidodendron, Lepidophloios))))))). Other parsimonious topologies allow dissociation of the Paurodendron-Oxroadia clade (probably justified), transposition or unification of Anabathra and Chaloneria, and addition of Sigillaria to the Diaphorodendron-Synchysidendron clade. The analysis confined to vegetative characters translocates Hizemodendron close to the base of the clade, thus uniting the non-trees as an ostensibly paraphyletic basal group. The analysis confined to reproductive characters more closely resembles the analysis of all characters, but fails to resolve relationships among the four basal, bisporangiate-coned genera, and between Hizemodendron and Lepidodendron. These observations cast doubt on the value of partial-plant and isolated-organ phylogenies. Parsimonious use of the increasingly sophisticated and K-selected reproductive strategies as the basis of the overall phylogeny inevitably renders homoplastic the partly discordant vegetative architectures (including the tree habit). Consequently, a poorly resolved paraphyletic plexus of four primitive, bisporangiate-coned genera (Paurodendron, Oxroadia, Anabathra, Chaloneria) subtends a monophyletic monosporangiate-coned clade of three well-supported, monophyletic families: the Sigillariaceae (Sigillaria) are primitive relative to the Diaphorodendraceae (Diaphorodendron sens. str., Synchysidendron) and the Lepidodendraceae (Hizemodendron, Lepidodendron sens. str., Lepidophloios), which together are characterized by a single functional megaspore per megasporangium. This apparently progressive evolutionary trend toward seedlike reproduction increased ecological specialization and is consistent with adaptive scenarios. In contrast with reproductive features, vegetative features such as the determinate growth, centralized rhizomorphic rootstock, and small number of module types that constitute the bauplan (rhizomorph and stem essential, lateral and crown branches optional) apparently predisposed the arborescent lycopsids to nonadaptive saltational evolution. Mutation of genes controlling early development allowed radical changes in growth architecture, and consequent epigenetic readjustment and adaptive honing affected many other vegetative characters. The progenetic (heterochronic) origin postulated for the pseudoherb Hizemodendron may also apply to Chaloneria and the other pseudoherbs (Paurodendron, Oxroadia), arguably comprising their value in scoring habit characters for the hypothetical ancestor. Other limitations of the present data matrix are the large number of genus-level autapomorphies (at least partly reflecting the absence of pre-Pennsylvanian arboreous species), the inclusion of only one bisporangiate-coned tree (Anabathra) and of only one putative isoetalean (Chaloneria). More primitive OTUs are needed to investigate the origins of profound character states shared by all OTUs in the present study (e.g., secondary thickening, determinate growth, centralized rhizomorph, heterospory), and to confirm the crucial hypotheses of monophyly for the monosporangiate cone and the single functional megaspore. Repeated simplification of growth architecture by progenesis and of megaspore ornamentation by functional redundancy show that evolution did not consistently increase morphological complexity among the arborescent lycopsids. Synapomorphies of highest burden (and therefore lowest homoplasy) tend to represent features of intermediate scale. We have not identified any significant drawbacks of cladistically analyzing an exclusively extinct set of OTUs. Rather, we recommend further study of some under-researched aspects of phylogeny reconstruction in general: (1) the effect of missing values on tree length calculations and on character state optimization; (2) the minimum acceptable level of empirical support (apomorphic states per OTU); (3) means of recognizing heterochrony in cladograms; and (4) less methodologically constrained phenetic adjuncts to strict cladistic analyses