Combined and Separate Analyses of Morphological and Molecular Data in the Plant Family Rubiaceae

Abstract

The Rubiaceae are one of the largest of the families of angiosperms, with over 10 000 species. The tribal and subfamilial classification is provisional due to the lack of phylogenetic hypotheses. The present study of the Rubiaceae is based on 33 genera and three data sets, one morphological and two molecular from chloroplast DNA, restriction sites and rbcL sequences. There is much congruence between the morphological and the molecular data sets, but also conflict. For parsimony reasons, the best phylogenetic hypothesis is a tree based on an analysis of the combined data sets. The so-called "total evidence" criterion for the combined analysis is simply a reiteration of the principle of parsimony. In this particular study, the classification would be almost the same even if based on the separate analyses instead of the combined. Despite the inapplicability of consensus trees or trees from separate analyses for phylogenetic hypotheses and classification, separate analyses may provide important information. It is the best way to reveal conflicts between different data sets. Knowledge of the conflicts can promote further detailed investigation in order to improve understanding of characters and phylogenetic hypotheses. In this study, the tribe Vanguerieae provides such an example; morphological data support a position in the subfamily Cinchonoideae, but DNA and a tree based on the combined data support a position in subfamily Ixoroideae. The tribe's position in the morphological tree is probably due to missing information concerning the correct pollen presentation system. Bootstrap fractions and K. Bremer's branch support values are used to evaluate the stability of particular nodes in the trees. Interestingly these values are not always correlated, e.g. in the morphological tree, the node with the highest branch support value has very low bootstrap fraction. The reasons for these differences are unclear, but large differences are presumably more likely to occur on short branches.

More Characters or More Taxa for a Robust Phylogeny—Case Study from the Coffee Family (Rubiaceae)

“…Eleven outgroups, representing the rest of the Gentianales or the sister group of Rubiaceae (Bremer, 1996b;Backlund et al, 1999), were used (Table 1). We also reanalyzed a set of Rubiaceae data from Bremer (1996a) : Rubiaceae 33 morph, Rubiaceae 33 RFLP, Rubiaceae 33 rbcL, and Rubiaceae 33-morph/RFLP/rbcL. DNA was extracted, ampli ed, and sequenced following the protocols of Bremer et al (1995) for rbcL and of Kim and Jansen (1995) or Backlund et al (1999) for ndhF.…”

“…Sequences are accessioned in EMBL (Table 1). A few sequences were obtained from GenBank, and the others have been published earlier by the rst author alone or in collaboration with others (Olmstead et al, 1993;Bremer et al, 1995;Bremer, 1996aBremer, , 1996bAndreasen and Bremer, 1996;Endress et al, 1996, Andreasen et al, 1998, and Backlund et al, 1999.…”

“…One debate concerned morphological versus molecular data (reviewed in Hillis, 1987;Patterson, 1987;Fernholm et al, 1989;Givnish and Sytsma, 1997). The consistency of the characters or the level of homoplasy has been shown to be about the same for morphological and molecular data (e.g., Sanderson and Donoghue [1989]; in Rubiaceae, Bremer [1996a]; and in Apocynaceae, Endress et al [1996]) . However, the number of characters is usually much higher in molecular data sets.…”

“…1b). The problem with morphological data is to get enough characters; in the Rubiaceae 33 morph analysis (Bremer, 1996a), the 35 characters are not suf cient to resolve the tree, or to give many supported nodes. For example, the morphological matrix should have at least 93 characters to provide 95% bootstrap values for all nodes-if the characters are binary, without homoplasy, and evenly distributed on the tree.…”

“…In recent years, phylogenetic analyses of representatives from all subfamilies and several outgroups based on different data sets (molecular and morphological) have been published (Bremer et al, 1995;Bremer, 1996a) and show that the family is monophyletic but that some of the deepest branches are unresolved or only weakly supported-a situation that makes subfamily classi cation uncertain and some taxa dif cult to classify. With this empirical study we hope to provide some new insights into the following questions: (1) Is it better to use more characters or more taxa to get a more robust (supported) phylogeny (at the base of the family tree and for the rest of the tree)?…”