PHYLOGENETIC RELATIONSHIPS IN GREEN PLANTS—A COMMENT ON THE USE OF 5S RIBOSOMAL RNA SEQUENCES BY BREMER ET AL.

Abstract

In a recent paper Bremer et al. (1987) presented a cladistic analysis of green plants using a variety of morphological, biochemical, and cytological features. They considered, but chose not to include, 5S rRNA sequence information in their data set. Prior to that decision they performed a parsimony analysis of 5S rRNA sequences presented by Hori et al. (1985); they illustrated one of the 57 most parsimonious trees from that analysis. A variety ofnovel are present in that tree. Based on the likelihood that those groups do not reflect the true relationships and on the extensive homoplasy of the 5S rRNA sequence data indicated in their analysis, Bremer et al. (1987) chose not to include those data in their phylogenetic reconstruction of green plants. In light of the increasing availability and use of both DNA and RNA sequence data we would like to comment on their dismissal of the 5S rRNA data. We will point out a few possible reasons for the unusual groupings produced in their analysis. Our consideration of 5S rRNA data demonstrates that one must take as much care with character analysis of those data as one does with morphological data. Our primary aim in this comment is to provide additional information on the use of RNA sequence data in order for others to make an informed decision on its use in phylogenetic reconstruction. At this time it is not our purpose to comment on specific phylogenetic hypotheses. We are in the process of reanalyzing the data presented by Hori et al. (1985) in view of the special features of rRNA discussed below. In preliminary analyses we have obtained trees with groupings congruent with current hypotheses on the phylogeny of green plants. The results of these analyses will be presented in a separate paper. Ribosomal RNA sequence data have been used in a variety of analyses to estimate phylogeny of plants, bacteria, and animals using both cladistic and phenetic methodologies (e.g., Hori et al., 1985; Hori and Osawa, 1986; Hixon and Brown, 1986). The analysis of the 5S rRNA sequence data by Hori et al. (1985) uses phenetic algorithms to infer the phylogeny of green plants. They use an equation derived by Kimura (1980) to calculate the evolutionary distance between any two sequences. These pair-wise distances were used to construct a hierarchical tree using the unweighted pair-group method using arithmetic averages (UPGMA) (Sneath and Sokal, 1973). Hori et al. (1985) use a phylogenetic interpretation of the branching patterns of the hierarchical tree and suggest for example, that the bryophytes evolved from ferns by degeneration. Cladistic analyses of sequence data using parsimony algorithms also have been performed, e.g., analyses by Hixon and Brown (1986) of 12S mitochondrial rRNA of primates. Their results, while not unequivocally supporting one hypothesis, do allow one to rule out some of the several phylogenetic hypotheses involving those taxa. These and other papers that use rRNA sequence data in phylogenetic analyses point out certain features that complicate its use for those purposes. Other workers have discussed these features in detail (Rothschild et al., 1986; Wheeler and Honeycutt, 1988). In the following discussion we will indicate how these features may be used in a character analysis that is essential for the use of sequence data in phylogenetic reconstruction. A minimal model of the secondary structure of the eucaryotic 5S rRNA molecule (Fig. 1) has portions that are single-stranded (loops) and portions that are double-stranded (stems) (Erdmann and Wolters, 1986). Nucleotide positions in the stem may not undergo independent base substitution. Substitution may occur in complementary bases in order to provide the Watson-Crick binding necessary for the functioning of the rRNA molecule. Wheeler and Honeycutt (1988) term this process cosubstitution. They examined the sequences of 5S RNA for 95 taxa and compared the rate of substitution in the loop versus double-stranded stem portions of the molecule. They found significantly greater rates of substitution in the stem regions than in the loop regions. They also found significantly increased rates of cosubstitution in the stem region compared with that expected from a binomial model. Wheeler and Honeycutt (1988) used structural information to formulate three data sets, one that includes all of the positions, one with only stem positions, and one with only loop positions. They used parsimony analysis to produce phylogenetic trees for each data set for a variety of taxa. Note that the first two data sets that include stem positions have redundant and thus effectively