Phylogenetic Relationships within North American Astragalus L. (Fabaceae)

Abstract

A cladistic analysis of 113 species and varieties of North American Astragalus is presented to develop hypotheses of phylogenetic relationship among sections and major clades. A set of 57 binary and unordered multistate characters derived from morphology, biochemistry, and cytology was analyzed using Wagner parsimony. Two large sets of equally parsimonious trees were found at 597 steps. Consensus and tree similarity methods were used to characterize each set and to compare them to each other. Features shared by both sets include a paraphyletic assemblage of sections at the base (Barneby's sects. Strigulosi, Micranthi, Miselli, Leptocarpi, Reventi-arrecti in part, and others), which gives rise to three large groups. The best supported clade is composed of sections having predominantly deciduous fruits, including sects. Inflati, Humistrati, Humillimi, Desperati, Drabellae, and Chaetodontes. Another large clade includes the bulk of Barneby's Homalobi, a group of taxa characterized by persistent, often laterally compressed fruits. The third group includes many sections with large flowers, such as sects. Argophylli, Sarcocarpi, Preussiani, and Trichopodi). In the first set of trees each of these groups is monophyletic; in the second set the Homalobi clade is derived from within the large-flowered clade, making the latter paraphyletic. Although uncertainty about the basal relationships of these three groups persists, other relationships are well supported. The impact of certain limitations in the analysis is discussed, including character/taxon ratio, sampling of taxa, method of rooting, and the exclusion of South American species from the study. The monophyly of sections and higher level relationships proposed by Rydberg, Jones, and Barneby is briefly evaluated in light of these results. Large genera such as Astragalus occupy a unique position among extant groups. Whether by accident of evolution or caprice of taxonomic history, they afford the opportunity to study evolutionary patterns replicated on a large scale. Important evolutionary processes may not be apparent in smaller genera. Features of character evolution including character correlation (Maddison 1990), coevolution (Sillen-Tullberg 1988), and adaptation (Coddington 1988) may be difficult to detect without the measure of statistical replication afforded by large taxonomic assemblages. Levels of homoplasy are higher in large taxa (Sanderson and Donoghue 1989), and patterns of parallelism and convergence are easier to detect there (Sanderson 1989a). However, reconstructing the phylogeny of the . . most difficult genus of North American plants (Jones 1923, p. 1), is not without problems. The large number of species (about 400 in North America), the uncertainty about relationships in the Old World, the lack of comprehensive data on characters other than gross morphology, the lack of any fossil record, all conspire against the successful elucidation of phylogenies of the group. Yet these difficulties are mitigated by several factors. The relationships of North American species have been studied in detail by four systematists since the middle of the last century (Gray 1864; Jones 1923; Rydberg 1929; Barneby 1964) and an updated species summary for the U.S. has recently been compiled (Isely 1983a, 1983b, 1984, 1985, 1986). Floristic work in the western U.S. has clarified the status of certain taxa (Barneby 1989; Spellenberg and Diaz 1988; Welsh 1978). Many species-level problems have been resolved because of this long-standing attention. Species distributions characterized by allopatry and local endemism with little hybridization (Barneby 1964) may imply that reticulate evolution has not been widespread. Cytology suggests the monophyly of the vast majority of the New World radiation, which (if ultimately supported) simplifies problems presented by uncertain relationships among Old World taxa. Previous taxonomic work in the genus emphasized the circumscription of sections rather than their relationships to each other (Barneby 1964; Jones 1923; Rydberg 1929). Testing the monophvlv of sections and uncovering larger