MAMMALIAN PHYLOGENY AND CONFLICTS BETWEEN MORPHOLOGICAL AND MOLECULAR-DATA

The phylogeny of mammals is an unresolved problem. Numerous conflictin g hypotheses on the interrelationships on the extant eutherian orders exist. The recent litterature emphasizes on one type of conflicts : co ntradiction between morphological and molecular data. In this paper we examine some of these conflicts and focus on two examples of differen t hierarchical levels : the interrelationships of the Hominoidea and t he relationships of the Tethytheria (elephants and sea-cows) to other eutherian orders or superorders. The question of the interrelationship s of the extant hominoid genera is analyzed through molecular and morp hological data. Parsimony analyses of the pseudo eta-globin gene of Ho mo, Pan, Gorilla, Pongo and two outgroups, Macaca and Ateles are run w ith PAUP (v. 3.0.). When characters are unweighted it gives only one t ree where Pan and Homo are sister groups (L = 209 steps, C.I. = 0.92). The pseudo eta-globin gene consists of 7236 sites, 311 being informat ive (and among these 311 sites, 100 are synapomorphies of the Hominoid ea). When transversions are weighted (weight = 2), or both transversio ns and gaps (insertions/deletions) are weighted (= 2) the clade (Homo, Pan) is not altered. It is only when gaps are suppressed (without wei ght on transitions and transversions) that the clades (Pan, Homo) and (Pan, Gorilla) are equally parsimonious. The morphological data (analy sis with Hennig 86) are based on 70 skeletal characters checked in the five living genera and in cercopithecoids and platyrrhines as outgrou ps. The most parsimonious tree (L = 105 steps, C.I. = 0.85) include th e clade (Pan, Gorilla). This example is used to test the importance of traits associated to << knuckle-walking >>. When the three assumed in dependant traits associated to knuckle-walking are reduced to one. two most parsimonious trees are obtained with the relationships [(Homo, P an) Gorilla] and [(Pan, Gorilla) Homo]. The discussion on the monophyl y of the Paenungulata (proboscideans, sirenians and hyracoids) is firs tly based on the separate parsimony analyses of alpha hemoglobin, beta hemoglobin, alpha + beta hemoglobin and alpha-A crystallin chains. Wi th alpha hemoglobin chain, we got two equally parsimonious trees, one of them with paenungulates not monophyletic and the other one with pae nungulates monophyletic. The analysis of the beta hemoglobin chain lea d to four equally parsimonious trees with a monophyly of Paenungulata- All of them differe from the two trees obtained with the alpha chain, With the tandemly combined alpha + beta chains, we obtain one parsimo nious tree who is one of the tree found with the beta chain. Though th e paenungulates are monopyletic in most analyses (except for alpha hem oglobin) the results are extremely flexible. For each analysis, the nu merous most parsimonious trees show conflicting evidence on the relati onships of other taxa (artiodactyls, cetaceans and perissodactyls) and among paenungulates. The review of the morphological data show also c onflicting results, though the more comprehensive study (77 unweighted anatomical character), does not support the monophyly of paenungulate s, and support sister group relationships of perissodactyls and hyraco ids. Clearly homoplasy affects all kind of data. Conflicts are either between morphological and molecular data, or among each kind of data. Among molecular studies the number of taxa seems particularly relevant if one considers the robustness of the hypotheses of relationships ba sed on each kind of proteins or genes. For morphological studies the a ddition of newly discovered characters is mainly responsible of change s in the phylogenetic hypotheses.