To understand the evolution of duplicate genes, we compared rates of nucleotide substitution between 17 pairs of nonallelic duplicated genes in the tetraploid frog Xenopus luevis with rates between the orthologous loci of human and rodent. For all duplicated X. laevis genes, the number of synonymous substitutions per site (ds) was greater than the number of nonsynonymous substitutions per site ( dN), indicating that these genes are subject to purifying selection. There was also a significant

more »

... itive correlation (Y = 0.9 15) between & for the X. laevis genes and & for the mammalian genes, suggesting that, at the amino acid level, the X. laevis genes and the mammalian genes are under similar constraints. Results of relativerate tests showed nearly equal rates of nonsynonymous substitution in each copy of the X. laevis genes; apparently there are similar constraints on both copies. No correlation was found between ds for the X. laevis genes and ds for the mammalian genes. There was a significant positive correlation both between members of pairs of duplicated X. laevis genes (Y = 0.95 1) and between human and rodent orthologues (r = 0.854) with respect to third-position G + C content but no such relationship between the X . laevis genes and either of their mammalian orthologues. The results indicate that both copies of a duplicate gene can be subject to purifying selection and thus support the hypothesis of selection against all genotypes containing a null allele at either of two duplicate loci. 1.