Detecting Evolutionary Incompatibilities From Protein Sequences

Abstract

Sneath, P. H. A., M. J. Sackin, and R. P. Ambler (Medical Research Council Microbial Systematics Unit, University of Leicester, Leicester LE1 7RH, England, and Department of Molecular Biology, University of Edinburgh, Edinburgh, EH9 3JR, Scotland) 1975. Detecting evolutionary incompatibilities from protein sequences. Syst. Zool. 24:311-332. The study of evolutionary phenomena such as reticulate evolution, gene transfer, convergence and mosaic evolution, and their differentiation from mistaken evolutionary homology (i.e., orthology), requires methods for detecting cladistic incompatibilities. Protein sequences are potentially valuable for such investigations, and for these two main strategies are proposed. Strategy (a) is to look for unexpected similarities between segments of the sequences using cross-association or similar methods. Strategy (b) (which assumes satisfactory prior alignment of the sequences) is to compare the sites in the sequence and test whether the observed co-occurrences would readily fit a single cladogram; for this Le Quesne's method of testing for cladistic incompatibilities has been adapted to protein sequences. Studies have been made on a number of cytochromes. Strategy (a) reveals unexpected high similarity among the N-terminal portions of some eukaryote and prokaryote cytochromes, and there appear to be several short segments that differ in their patterns of similarity. It is not yet clear whether functional requirements or gene transfer is the explanation. Strategy (b) has been tested on sequences where the evolutionary history appears straightforward and also on simulated disturbances of cladogeny and homologization. It seems to be a sensitive technique for such studies. Strategy (b) has also the advantage that it should permit the partitioning of characters into sets that may reflect different evolutionary behavior even when such sets cannot be distinguished by inspection. Preliminary trials suggest that such sets could be recovered by cluster analysis of character compatibility matrices. [Evolutionary Incompatibilities; Protein sequences.] There is a growing interest in detecting and distinguishing such evolutionary phenomena as gene transfer, convergence, mosaic evolution and mistaken phylogenetic homologies (Farris, 1971; Sneath and Sokal, 1973; Sneath, 1974). The methods proposed for such studies usually require the comparison between relationships based upon different sets of characters. Farris (1971) has discussed at some length the question of cladistic incompatibility between character sets as defined by Camin and Sokal (1965), that is the phenomenon whereby different sets of characters imply different topocladograms (cladograms recording only topology: Sneath, 1974). He has noted that mosaic evolution, i.e. faster evolution in some sets of characters than in others (De Beer, 1954), causes incongruence in the phenetic similarities implied by such sets-in other words it causes phenetic incompatibilities, and results in different orthocladograms (cladograms with minimum evolutionary distances associated with each internode; Sneath, 1974). In principle, therefore, one can distinguish mosaic evolution from homoplasia by methods suggested by Farris. However, it has been pointed out (Sneath, 1974) that other evolutionary phenomena (which we will here treat as exceptional ones)-gene transfer and mistaken phylogenetic homology (orthology in the sense of Fitch, 1970)-can also lead to phenetic or cladistic incompatibilities. The exceptional phenomena can in principle be distinguished, provided that the evolution in those parts of the phylogeny that are not involved in the exceptional phenomenon has been substantially constant and