Correlation between genome size, observed codon preference, and Gibbs energy of codon-anticodon interaction

The universal correspondence between the primary structure of a gene, characterized by a unique sequence of only four different nucleotide residues, and the primary structure of a protein, characterized by a given sequence of twenty different amino acid residues, is known as the "genetic code". This code is degenerated in the sense that a group of six, four or just two synonymous codons can code for a given amino acid. In this paper I will discuss only the special features of duet codons and in particular the effect of the purine/pyrimidine base proper in the third position, the observed substitution frequency, and the major determinant of the codon usage bias in prokaryots and in eukaryots. INTRODUCTION Large scale sequencing of DNA from a variety of sources has led to a new insight into the bias affecting the observed frequency of any of the 64 possible codons. We can still hold that the classic genetic code can be viewed as the universal language of all living beings. (1) It is obvious, however, from our knowledge as it stands today that many species speak their own dialect. Distinct codon strategies were found for different genomes. Yet a surprising consistency of choices was found among codons in genes of the same or of related genomes. (2) Within a given genome codon usage is conserved in most genes. This finding has been confmed for numerous organisms. E. coli genes, coding for proteins with a wide variety of functions have now been sequenced, enclosing abundant proteins for ribosomes etc., as well as for proteins that are present in only a few copies per cell, such as lac repressor. There are two important points to make: First, there are similarities in codon choices among all different E. coli genes (3); Second, although all E. coli genes show this tendency to confm to a specific codon usage pattern, there is a clear difference in the preferred codon set between highly and poorly expressed genes. (4) It was found for many E. coli genes that the degree of bias in codon choice is directly related to the expression level of a particular gene. Using the codon usage data of these genes allows to deduce a set of favoured codons. It is interesting to note that the mean GC content (the percentage of G/C bases in either the fist, second or third position of a set of codons for the 20 amino acids) of this biased set matches perfectly with the mean G/C content of the whole genome. (50 % /50% for E. coli). (5 ) It can be shown that this general rule holds for other prokaryots such as B. subtilis or T. ferrooxidans i.a. as well. Based on this rule one can propose a preferred for any prokaryot when the mean value of the G/C content of the genome is known. Synonymous codon choice patterns in other enterobacteriacaea such as S. typhimuriuni, K. pneumoniae, or Erwinia amylovora were found to resemble that of E. coli, confirming an earlier finding of Grantham who stated for prokaryots that taxonomically related organisms exhibit a related codon preference. It was shown in the mean time that for eukaryots with a small genome size such as S. cerevisiae a similar relation between the G/C content (6) of the favoured codon set and the G/C content of the total genome applies. Again choices among synonymous codons are strongly biased, and a clear relationship between the degree of bias and the level of expression of the gene proper has been observed. The extend of this bias exceeds that found in E. coli. A comparison between the preferred codons for yeast with those in E. coli shows that these two organisms use completely different codons for five amino acids (glu, lys, pro, leu, and arg). (2) The study of codon usage in higher eukaryots is complicated by the cellular differentiation. Examining the genes which code for the immunoglobulins, globulins, or peptide hormones it was found that mammalian genes use a codon strategy that must be very different from both bacteria and yeast genes. In general, A ending codons are avoided, Cand Gending codons are predominant for twofold degenerated codons e.g. Various laboratories have examined codon usage in vertebrate muscle genes (avian alpha actin, myosine lighdheavy chain i.a. and liver genes (rat albumin etc.) and have found a considerable degree of similarity of synonymous codon usage between the two groups of genes. Viewing the data of a large variety of genes from higher eukaryots one can conclude that for vertebrates a consensus codon usage pattern exists which is independent of taxonomic class (7) and tissues of expression. It should be noted, however, that minor but nevertheless important differences in codon usage can exist between highly and poorly expressed genes. It should also be added here that as in the case of E. coli and its phages the codon strategies of mammalian viruses (SV40, adenovirus, hepatitis v,) differ from those of their hosts. It is noteworthy that in all higher vertebrates NCG codons (N=A, T,C and G) in the serine, 1947