The evolution of strand-specific compositional bias. A case study in the Hymenopteran mitochondrial 16S rRNA gene.

Nucleotide sequences from a variety of sources display compositional bias (e.g., Bernardi et al. 1985; Jukes and Bhushan 1986; Sueoka 1988; Crozier and Crozier 1993; Dowton and Austin 1995). This bias in the proportion of the four nucleotides in a particular DNA fragment was originally thought to arise from mutational bias in the interconversion of adenine-thymine (AT) and guanine-cytosine (GC) base pairs (Sueoka 1962). However, the mechanism whereby compositional bias develops must be more complex than this. Two distinct types of compositional bias are observed: strand-specific and strand-nonspecific. Strand-specific compositional bias occurs when one strand contains higher levels of certain nucleotides over others, e.g., A+C or A+G bias. It is thought to develop when the rate of nucleotide substitution (e.g., A+C) differs from that in the opposite direction (e.g., C+A) (Jermiin et al. 1996). Strand-nonspecific compositional bias occurs when both strands contain a high level of A+T or G+C. Trends in compositional bias may follow phylogenetic lineages (Sueoka 1962; Hori and Osawa 1987; Jermiin and Crozier 1994; Jermiin et al. 1994; Dowton and Austin 1995). This suggests that a phylogenetic analysis of compositional bias may lead to an understanding of how this bias develops. In the Hymenoptera (wasps, bees, ants), broad phylogenetic relationships have been consistently suggested from fossil (Rasnitsyn 1980, 1988), morphological (Gibson 1985), and molecular (Dowton and Austin 1994) data (see fig. 1). We recently traced an increase in compositional bias to one particular hymenopteran lineage, the Apocrita (Dowton and Austin 1995). The Hymenoptera thus represent an ideal model group in which to study the evolution of compositional bias. In the present study, we determined the individual nucleotide contents of a fragment of the 16s rRNA gene, homologous to nucleotides 13474-13895 of the honeybee mitochondrial genome (Crozier and Crozier 1993). We used all available hymenopteran homologues at the time of writing, except for one wasp, Schlettererius cinctipes. This taxon was omitted because of its very different compositional bias and greatly increased rate of evolution compared with the included wasps (Dowton and Austin 1995). Together with its misplacement in a hymenopteran molecular phylogeny based on