The origin and relationships of flowering plants are among evolutionary biology's enduring mysteries. These comparative newcomers to the evolutionary stage number a staggering 250,000 living species classified into about 350 families. One of the cherished goals of botany is to unravel the complicated family tree, a process that involves establishing the branching relationships among species (phylogeny). Two papers in this issue take an impressive step forward. Soltis et al. (page 402)1 and Qiu et al. (page 404)2 use a multigene approach to identify the closest living relatives of flowering plants and map out the deepest branches of the family tree.
Analysing the deep phylogenetic relationships in flowering plants is a complicated business. This is partly because of the sheer size and diversity of the group, and partly because of its rapid diversification during the Early Cretaceous (130–90 million years ago)3. Flowering plants differ considerably from their closest living relatives in the gymnosperms (conifers and their allies), creating additional problems for 'rooting' the family tree. Recent molecular systematic work — which includes some of the largest phylogenetic analyses attempted for any group of organisms — confirms the existence of a major group (eudicots) which has a characteristic, three-aperture pollen type and includes most of the dicotyledons4, 5. The widely recognized division between dicotyledons and monocotyledons based principally on embryology (number of seed leaves), leaf morphology (veination type) and floral organization (number of parts) is not upheld. Furthermore, it is now clear that several of the traditionally recognized subclasses, such as the Dilleniidae and Hamamelidae, are not natural groups.
As to the earliest branches of the flowering-plant tree, there is strong evidence that these lie within the so-called magnoliid dicotyledons, a small group (only about 3% of living species diversity) which includes such familiar forms as water lilies (Nymphaeales), true peppers (Piperales), magnolias and several other related families such as the Winteraceae, Chloranthaceae and Lauraceae. But despite these and other successes, the exact pattern of branching among basal lineages has remained obscure.
One weakness of current approaches is that they are built mainly on single-gene data sets. Soltis et al. and Qiu et al. rectify things by combining gene sequences from different genomes (nucleus, mitochondrion, chloroplast) in individual species. As well as increasing the total data, this strategy is designed to reduce noise generated by gene-, function- and genome-specific phenomena. The results have broad implications for our understanding of flowering-plant evolution.
The two papers take slightly different approaches. Qiu et al.2 amass a huge sequence database (five genes; 8,733 base pairs) for a comparatively modest number (105 species) of exemplar organisms; these are species selected to represent larger groups. Soltis et al.1 sacrifice sequence length (three genes; 4,733 base pairs) for breadth of taxonomic coverage (560 species).
In their analysis, Qiu et al. provide what is arguably the first really solid evidence for the basal three branches of the flowering-plant tree (Fig. 1). They show that the family Amborellaceae, which consists of the single species Amborella trichopoda, is the sister group to all other flowering plants, a result that is consistent with another analysis based on duplicate phytochrome genes6. Next come the water lilies (Nymphaeales), followed by a group dubbed ITA (Illiciaceae, Trimeniaceae, Austrobaileyaceae, Schisandraceae). It is the strength of evidence on these branches that really distinguishes this result from other large analyses based on fewer genes. It really looks as though Qiu et al. might have got the answer to the seemingly intractable problem of basal lineages in flowering plants.
Figure 1: Summary of relationships among the major groups of flowering plants and gymnosperms based on the multigene sequence analysis of Qiu et al.2.
Thicker branches are those with strongest (90–100% bootstrap) support from the analysis. In this scheme, dicotyledons are not recognized as a group within the flowering plants.
High resolution image and legend (51K)How does the approach of Soltis et al. perform on this question? The tree topology is identical, but branch support for basal lineages is stronger in the study of Qiu et al.. This would seem to vindicate Qiu and colleagues' approach: at this level it is not so much adding more taxa that yields better supported results but having more data from a comprehensive selection of exemplars. One must however be cautious with the exemplar approach, because it can lead to blind spots. Amborella, for example, did not appear in some earlier analyses because it was not a first-choice exemplar, and it is inconveniently confined to the Pacific island of New Caledonia!
The interest of the paper by Soltis et al. is in the breadth of taxonomic coverage. The general pattern of relationships is similar to that outlined by Qiu et al., but there is more detail within the monocotyledons and eudicots. The relationships of model organisms and crops — such as Arabidopsis thaliana (thale cress), Nicotiana tabacum (tobacco), Brassica oleracea (cabbage, kale) and Solanum tuberosum (potato) — are defined more precisely. This information is essential for formulating general principles about the biology of flowering plants from particular experimental observations. Among other things, the analysis strongly supports a close relationship between families of nodule-forming nitrogen-fixing plants and the multiple origins of petals from sepals. Here we indeed have the beginnings of a general phylogenetic framework for flowering plants — a powerful research tool for comparative biology.
One surprising outcome of Qiu and colleagues' analysis is that it refutes the widely held 'anthophyte hypothesis' for the origin of flowering plants. This hypothesis links flowering plants to a small and obscure group of living gymnosperms called Gnetales and the extinct Bennettitales, based on morphological similarities7. The molecular data suggest otherwise. Qiu et al. provide strong evidence that Gnetales group within the conifers to the exclusion of flowering plants. If this shift in relations holds up, it will have far-reaching consequences for our understanding of the evolution of such flowering-plant characteristics as the carpel, double fertilization and the flower itself7, 8. Furthermore, it implies that the split between the flowering-plant lineage and other living gymnosperms occurred much earlier than currently supposed, possibly by the end of the Carboniferous, 290 million years ago.
