Phylogenetics of the Cotton Genus (Gossypium): Character-State Weighted Parsimony Analysis of Chloroplast-DNA Restriction Site Data and Its Systematic and Biogeographic Implications

ABsTRAcr. Total genomic DNAs from 61 accessions of 40 species of Gossypium were surveyed for restriction site variation in the maternally inherited plastid genome using 25 endonucleases. One hundred thirty-five of the 202 restriction site variants detected were potentially synapomorphous and served as binary characters for phylogeny estimation. Two cladistic methods were employed: Wagner parsimony analysis, which resulted in four equally most-parsimonious topologies requiring 161 steps (CI = 0.84), and a novel character-state weighting approach that models the relative probabilities of restriction site losses versus gains. This latter technique, which is theoretically preferable to both Dollo and Wagner parsimony analysis in that it optimizes against parallel site gains, resulted in two optimal phylogenetic estimates (a subset of the Wagner topologies) that differ only in the placement of G. longicalyx. In general, maternal cladistic relationships are congruent with both cytogenetic groupings and geographic clustering. Three major monophyletic clades among diploid species correspond to three continents: Australia (C-, G-genome), the Americas (D-genome), and Africa (A-, E-, and F-genome). African B-genome diploids are placed as sister to the New World D-genome species, albeit by a single homoplasious character state. Substantial agreement is also evident between the cpDNA phylogeny and traditional taxonomic treatments, although there is considerable disagreement at lower infrageneric ranks, particularly among the American and Australian cottons. These discrepancies are discussed, as is the possibility that inconsistency may reflect, at least in part, reticulation events among diploids, which may have occurred in at least three cases. An area cladogram suggests that Gossypium originated in either Africa or Australia. Because paleocontinental reconstructions, palynological evidence and cpDNA sequence divergence estimates concur in suggesting that the two primary clades diverged during the mid to upper Oligocene, the initial cladogenetic event most likely involved long-distance, intercontinental dispersal. Two colonizations of the New World are indicated, a relatively early long-distance dispersal from Africa leading to the evolution of the D-genome diploids, and a second, later dispersal of the maternal, A-genome ancestor of the allopolyploids. American diploid species are hypothesized to have originated in northwestern Mexico, with later radiations into other regions. The radiation of Gossypium in Australia is suggested to have proceeded from the westernmost portion of the continent. The maternal phylogenetic hypothesis and area cladogram suggest the possibility that New World allopolyploids originated following a trans-Pacific transfer of an ancestral A-genome taxon to the Pacific coast of Mesoamerica or South America.