Iridaceae ‘Out of Australasia’? Phylogeny, Biogeography, and Divergence Time Based on Plastid DNA Sequences

The current infrafamilial taxonomy of the Iridaceae recognizes four subfamilies; Isophysidoideae (1: 1); Nivenioideae (6: ca. 92), Iridoideae (29: 890), and Crocoideae (29: 1032). Phylogenetic analyses of sequences of five plastid DNA regions, rbcL, rps4, trnL–F, matK, and rps16, confirm most aspects of this classification and the evolutionary patterns that they imply, importantly the sisiter relationship of Isophysidoideae to the remainder of the family and the monophyly of Iridoideae. Subfamily Nivenioideae is, however, paraphyletic; Crocoideae is consistently found nested within it, sister to the core Nivenioideae, the woody Klattia, Nivenia, and Witsenia. This clade is sister to Aristea, which in turn is sister to the Madagascan Geosiris, and then to the Australasian Patersonia. We treat Aristea, Geosiris, and Patersonia as separate subfamilies, Aristeoideae and the new Geosiridaceae and Patersonioideae, rendering Nivenioideae and Crocoideae monophyletic. The alternative, uniting a widely circumscribed Nivenioideae and Crocoideae, seems undesirable because Nivenioideae have none of the numerous synapomorphies of Crocoideae, and that subfamily includes more than half the total species of Iridaceae. Main synapomorphies of Crocoideae are: pollen operculate; exine perforate; ovule campylotropous; root xylem vessels with simple perforations; rootstock a corm; inflorescence usually a spike; plants deciduous. Four more derived features of Crocoideae are shared only with core Nivenioideae: flowers long-lived; perianth tube well developed; flowers sessile; and septal nectaries present. The genera of the latter subfamily are evergreen shrubs, have monocot-type secondary growth, tangentially flattened seeds, and the inflorescence unit is a binate rhipidium. The latter feature unites core Nivenioideae with Aristea, Geosiris, and Patersonia, which have fugaceous flowers and, with few exceptions, a blue perianth. Molecularbased phylogenetic trees using sequences from five plastid DNA regions now show discrete generic clusters within Crocoideae and Iridoideae, the foundation for the tribal classification. The five tribe classification of Iridoideae, initially based on morphological characters and subsequently supported by a four plastid DNA region sequence analysis, continues to receive support using additional DNA sequences. Application of molecular clock techniques to our phylogeny indicates that the Iridaceae differentiated in the late Cretaceous and diverged from the next most closely related family, Doryanthaceae circa 82 mya, thus during the Campanian. The Tasmanian Isophysis is the only extant member of the clade sister to the remainder of the Iridaceae, from which it may have diverged 66 mya, in the Maastrichtian. The generic phylogeny shows the proximal clades of the family are all Australasian, which corroborates past hypotheses that the Iridaceae originated in Antarctica-Australasia, although its subsequent radiation occurred elsewhere, notably in southern Africa and temperate and highland South America at the end of the Eocene or later. Keywords—divergence times, Geosiridoideae, infrafamilial classification, new subfamilies, plastid DNA sequences, Patersonioideae. With over 2030 species divided among 65–75 genera (Table 1), Iridaceae are among the largest families of the order Asparagales (Goldblatt 2001a). In the first attempt to understand the phylogeny of the family since cladistic methods have been a focus of plant systematics, Goldblatt (1990a), using 52 traditional characters, identified four main clades corresponding to major generic clusters often recognized in classifications of the family as subfamilies or tribes. In the resulting classification, these clades were recognized as subfamilies: Isophysidoideae (1 genus: 1 species); Nivenioideae (6: 92); Iridoideae (20: ca. 890); and Crocoideae (then called Ixioideae) (29: 1032). The study was flawed in that characters were not always polarized by immediate outgroup comparison. However, no one at that time could have selected a suitable outgroup from among the many families that have been thought to be allied to Iridaceae. Indeed, the association of Iridaceae with the Asparagales was then novel (Dahlgren et al. 1985) and not widely accepted. A second study using 33 morphological and anatomical characters (Rudall 1994, 1995), placing more emphasis on anatomical features, reached somewhat different results. Genera of Crocoideae were sister to the rest of the family which divided into two clusters of genera, one corresponding to Iridoideae and the other including the genera of Goldblatt’s Nivenioideae plus Isophysis. This was clearly a different topology, although with more or less identical generic clusters in the major clades, but notably with Isophysis no longer sister to the remaining genera of the family. Molecular techniques were first applied to the family in a study by Souza-Chies et al. (1997) using the plastid gene rps4. Their results confirmed the monophyly of Iridaceae and supported the position of Isophysis, only member of Isophysidoideae, as sister to the remainder of the family. Their conclusion is tempered by an unfortunate choice of outgroup, genera of Amaryllidaceae and Agavaceae, families not immediately allied to Iridaceae, and this may have affected the resulting topology within the family. Subsequently, Reeves et al. (2001), in the first multigene study of the phylogeny of the family, and using sequences of four plastid DNA regions, rbcL, rps4, the trnL intron, and the trnL–F intergene spacer, reached conclusions broadly congruent with Goldblatt’s and Souza-Chies et al.’s results. By this time outgroups to the Iridaceae were known to comprise Doryanthaceae, Ixiolirion (Ixioliriaceae), and Tecophilaeaceae, in that order (Fay et al. 2000; Graham et al. 2006; Pires et al. 2006). While the exact relationships of the families in this cluster have still not been satisfactorily resolved, they comprise an appropriate outgroup. Reeves et al.’s study made it clear that Goldblatt’s (1990a) morphology-based conclusions were largely correct. These conclusions were: 1) that Systematic Botany (2008), 33(3): pp. 495–508 © Copyright 2008 by the American Society of Plant Taxonomists