Mispairing and compensational changes during the evolution of mitochondrial ribosomal RNA.

Understanding rates and patterns of ribosomal RNA (rRNA) nucleotide change is important for understanding RNA structure/function relationships ( Brimacombe et al. 1986; Stem et al. 1988), for predicting higher-order rRNA tertiary and RNAprotein interactions (Zuker and Stiegler 198 1; Freier et al. 1986 ), for studies of rRNA molecular evolution (Gerbi 1985; Gutell et al. 1985; Rousset et al. 1991), and for accurate phylogenetic analysis based on rRNA primary sequences (Wheeler and Honeycutt 1988; Smith 1989). rRNA secondary structure is highly conserved across widely divergent taxa, despite considerable variation at the primary-sequence level (Gerbi 1985; Gutell et al. 1985; Engberg et al. 1990). Within a particular rRNA gene, variation in the rate of primary-sequence evolution exists, with single-stranded loops evolving more rapidly than do presumed double-stranded stems (Tanhauser 1985). This result is expected if the complementary base for a particular stem position exerts some constraint on the allowable evolution in its opposite, paired member. Such constraint has been demonstrated for the 5s rRNA gene, where evolution in the stem regions is highly directed (Wheeler and Honeycutt 1988). Constraint on the evolutionary freedom of rRNA stem positions may be assessed by investigating the degree to which changes in one nucleotide of a stem base pair are compensated in the other member of the base pair (Wheeler and Honeycutt 1988). In addition, characterizing persistent mispairings may provide some insights into selective pressures on rRNA molecules. Here we present such an investigation for the 12s rRNA genes of the mitochondrial DNA (mtDNA) from an assortment of artiodactyls. This order of eutherian mammals has a diversity of extant and fossil species, thereby presenting a series of relatively well-dated cladogenetic events covering a large time range and several hierarchical taxonomic levels (Miyamoto et al., accepted). The tree topology and times of divergence used to study the evolution of artiodactyl 12s rRNAs are shown in figure 1. The topology for bovines (Miyamoto et al. 1989), cervids (Miyamoto et al. 1990)) and higher-level pecoran relationships (Simpson 1945; Janis and Scott 1987; Gentry and Hooker 1988) is based only on unambiguously well-diagnosed lineages. Dates of hypothetical common ancestors are those justified previously by Miyamoto et al. (accepted). The sequences themselves are from previous studies (Anderson et al. 1982; Tanhauser 1985; Miyamoto et al. 1990; Kraus and Miyamoto 199 1) . Fifty-one stems were identified by using the secondary-structure model of Gutell et al. ( 1985), except that eight pairs of stems, separated by a single base pair in that model, were each joined together to form eight composite stems, for a total of 43 stems. Twenty-nine of these stems contained 79 pairs of nucleotides that were variable in at least one taxon. We optimized these 79 nucleotide pairs on the tree of figure 1 by using the computer program PAUP [ Phylogenetic Analysis Using Parsimony, ver-