Codon Usage Bias, tRNA Modifications and Translational Selection in Bacteria

Abstract

<p>In this thesis, codon usage bias is studied in a large number of bacterial genomes. The codon frequency of each codon is measured as a fraction of the total number of codons for each amino acid, which can be used to calculate the potential selection strength in a genome according to the population genetics theory of selection, mutation and drift. It is assumed that codon frequencies in low expression genes are affected principally by mutation rates, but the frequencies in high expression genes are controlled by both selection and mutation. The ribosomal protein genes and elongation factors are used as high expression genes, while the rest as low expression genes. By comparing the frequencies of codons in high and low expression genes, we can determine the strength of selection acting on high expression genes. A model of translation kinetics is developed, which predicts the way that the speed of translation of each codon depends on the number of copies of each type of tRNA gene in the genome. This theory reveals that codon usage and tRNA gene copy number have to co-adapt to each other to optimize the speed of translation. We show that there are often multiple possible stable combinations of tRNAs and codon usage. This explains the observation that different codons are sometimes preferred for different amino acids in the same organism.</p> <p>We compare our theory with observed codon frequencies in a large number of bacterial genomes. Codon families are grouped according to different combinations of tRNAs and the averages of codon frequencies for all combinations are taken respectively. A preferred codon is defined as the codon whose frequency increases from low expression genes to high expression genes. The interesting phenomenon is that the anticodon-codon interactions are different between two-codon families (the U+C and A+G codon families) and four-codon families, mainly because of the tRNA modifications on position 34 of the tRNA anticodon loop. We show that the preferred codon depends on which base is present at wobble position and whether a base modification happens. It is found that the preferred anticodon-codon combinations do not always correspond to Watson-Crick pairs. In particular, in four-codon families, tRNAs with U at the wobble position appear to interact surprisingly well with U-ending codons.</p> <p>We introduce a parameter, K, which measures the strength of codon bias in each genome. We show that K is strongly correlated with the growth rate of the organism. This is consistent with the idea that selection for translational speed is most important for rapidly multiplying organisms. From the theory, the effective population size Ne is also expected to influence the selection strength. However, our data analysis shows that K is not correlated with Neu. This shows that variation of Neu among species is not a confounding factor in our interpretation. Although we believe that translation speed is the main reason for selection on codon usage in bacteria, it is also possible that selection for translational accuracy plays a role. Selection for accuracy can be tested by comparing codon frequencies in conserved and variable sites within the same gene sequences. We introduce a new statistical test to measure the strength of this effect. We observe a small but significant effect of accuracy with this method.</p>