Open Collections

Abstract

Genes are expected to face stronger selective constraint and to evolve more slowly if they encode enzymes upstream as opposed to downstream in metabolic pathways, since upstream genes are more pleiotropic, being required for a wider range of end-products. However, few clear examples of this trend in evolutionary rate variation exist. We examined whether genes involved in plant terpenoid biosynthesis exhibit such a pattern, using data for 45 genes from four fullysequenced angiosperms, Vitis, Arabidopsis, Populus and Ricinus. Taking a bioinformatic approach, we used phylogenetic trees of proteins from each genome to determine orthologs and count paralogs in each genome, using proteins from the Oryza genome as an outgroup. Our results show that dN/dS does in fact correlate with pathway position along pathways converting glucose to the terpenoid phytohormones abscissic acid, gibberellic acid and brassinosteroids. Upstream versus downstream rate variation is particularly strong in the gibberellic acid pathway, but absent in the pathway to lutein, another terpenoid derivative. In contrast, we found no apparent variation of dN/dS with gene copy number. We also introduce a new measure of pathway position, the Pathway Pleiotropy Index (PPI), which counts groups of enzymes between pathway branchpoints. We found that this measure is superior to pathway position in explaining variation in dN/dS along each pathway. Kendall's τ for the correlation of PPI with dN/dS was 0.24 - 0.54 with a mean of 0.4. We further show that variation in dN/dS is due to differences in selective constraint, not positive selection. Therefore, our results are consistent with the prediction that selective constraint is progressively relaxed along metabolic pathways, showing plant terpenoid synthesis to be a robust example of positional rate variation.