Pruned Median Networks: A Technique for Reducing the Complexity of Median Networks Katharina T. Huber,* Vincent Moulton,* Peter Lockhart,² and AndreasDress ‡ * FMI (Physics and Mathematics Department), Mid Sweden University, S 851-70 Sundsvall, Sweden; ² Institute of Biomolecular Sciences, Massey University, PO Box 11-222, Palmerston North, New Zealand; and ‡FSPM-Strukturbildungsprozesse, University of Bielefeld, D-33501 Bielefeld, Germany Received October 22, 2000; revised January 17, 2001 Observations from molecular marker studies on re- cently diverged species indicate that substitution pat- terns in DNA sequences can often be complex and poorly described by tree-like bifurcating evolutionary models. These observations might result from pro- cesses of species diversification and/or processes of sequence evolution that are not tree-like. In these cases, bifurcating tree representations provide poor visualization of phylogenetic signals in sequence data. In this paper, we use median networks to study DNA sequence substitution patterns in plant nuclear and chloroplast markers. We describe how to prune me- dian networks to obtain so called pruned median net- works. These simpler networks may help to provide a useful framework for investigating the phylogenetic complexity of recently diverged taxa with hybrid ori- gins. © 2001 Academic Press Key Words: phylogenetic network; Buneman graph; lite Buneman graph; median network; pruned median network; split; buttercups; Ranunculus; ITS markers; chloroplast JSA markers. INTRODUCTION Phylogenetic networks provide an alternative ap- proach to bifurcating tree models for representing phy- logenetic relationships and are constructed from “splits” that are either observed or calculated from the data [see Bandelt (1991, 1995), Dress et al. (1996), and Huson (1998) for examples]. They allow a visualization of the signals in data even when these reflect contra- dictory or complex evolutionary histories. In contrast to standard implementations of parsimony, neighbor- joining, or maximum-likelihood in evolutionary tree building, they do not require that the evolutionary process be perceived as a pure branching process and hence be modeled by a unique bifurcating tree. Median networks, which are typically constructed from binary character state data, are based on the concepts of compatibility and representation within hypercubes. They have already been used successfully to analyze population data in the form of human mito- chondrial sequences [see Bandelt (1995) and Watson et al. (1997) for examples], but they have rarely been used to examine the relationships between species. How- ever, since hybridization (Crawford and Stuessy, 1997; Wolfe et al., 1998) and little genetic change (Knox and Palmer, 1995; Crawford and Stuessy, 1997; Schaal et al., 1998; Ungerer et al., 1998; Winkworth et al., 1999) can accompany rapid morphological diversification of plant species, median graphs may also prove useful in such a context. In plant studies, patterns of sequence variation have been considered evidence for the hybrid origin of plant genomes. With nuclear DNA sequences, homologues originating from different parental species in hybrid taxa can sometimes be identified by the presence of “heteroplasminc sites” (Sang et al., 1995; Aguilar et al., 1999). These are sites at which more than one charac- ter state is present. However, interpretation of the information at these sites may also be complicated by processes of concerted evolution which can completely or partially convert one parental sequence into another (Buckler et al., 1997; Aguilar et al., 1999). In this case, evidence for the hybrid nature of a taxon may be a recombinant sequence without heteroplasmic sites. DNA sequences from cytoplasmic genomes (mitochon- drial and chloroplast DNAs) may also potentially pro- vide evidence for hybrid origins. If these genomes un- dergo intermolecular recombination as controversially suggested for human mitochondrial DNA (Macaulay et al., 1999; Eyre-Walker et al., 1999; Awadalla et al., 1999) and discussed for plant chloroplast DNAs (Howe, 1986), then such recombinant sequences exhibiting patterns of incompatibility could result. These pat- terns would not be described by bifurcating evolution- ary tree models. In this paper, we use median networks to represent the information in nuclear and chloroplast DNA se- quences from alpine and lowland species of Ranuncu- lus. We also introduce a new procedure for pruning Molecular Phylogenetics and Evolution Vol. 19, No. 2, May, pp. 302–310, 2001 doi:10.1006/mpev.2001.0935, available online at http://www.idealibrary.com on 1055-7903/01 $35.00 Copyright © 2001 by Academic Press All rights of reproduction in any form reserved. 302