Synonymous codon usage bias in Oryza sativa

By using 1860 cDNA sequences containing complete coding sequence (CDS), a detailed analysis of the effects of several factors on codon usage in rice was performed. The results showed that axis 1 coordinates were significantly positively correlated with GC3s and GC content (r = 0.993 and 0.954, respectively, P < 0.01), which indicated that nucleotide compositional mutation bias is the major factor in shaping the codon usage of rice. From the results of a lower but significant positive correlation between axis 1 coordinates and gene expression level assessed by Codon Adaptation Index (CAI) (r = 0.743, P < 0.01), and lower correlations between CAI values and Effective Number of Codons (ENC) values, GC3s, GC content (r = -0.717, 0.757 and 0.626, respectively, P < 0.01), we could conclude that genes with higher expression level showed more significant variation in codon usage, and they were strongly preferential for codons with C or G at the third nucleotide position. There is significantly positive correlation between CDS length and ENC values (r = 0.280, P < 0.01), but negative correlations between CDS length and axis 1 coordinates, CAI values (r = -0.322 and -0.316, respectively, P < 0.01), suggesting a tendency to higher levels of expression for genes with shorter CDS, and to lower levels of expression for genes with longer CDS at least at the translational level. In summary, the base composition of genes and the natural selection on gene expression level play a crucial role, and CDS length only in a minor way, in shaping codon usage of rice. In this paper, twenty-six codons defined firstly as optimal codons may provide some more useful information for rice gene-transformation. (C) 2004 Elsevier Ireland Ltd. All rights reserved.