A four-column theory for the origin of the genetic code: tracing the evolutionary pathways that gave rise to an optimized code.

backgroundThe arrangement of the amino acids in the genetic code is such that neighbouring codons are assigned to amino acids with similar physical properties. Hence, the effects of translational error are minimized with respect to randomly reshuffled codes. Further inspection reveals that it is amino acids in the same column of the code (i.e. same second base) that are similar, whereas those in the same row show no particular similarity. We propose a 'four-column' theory for the origin of the code that explains how the action of selection during the build-up of the code leads to a final code that has the observed properties.

resultsThe theory makes the following propositions. (i) The earliest amino acids in the code were those that are easiest to synthesize non-biologically, namely Gly, Ala, Asp, Glu and Val. (ii) These amino acids are assigned to codons with G at first position. Therefore the first code may have used only these codons. (iii) The code rapidly developed into a four-column code where all codons in the same column coded for the same amino acid: NUN = Val, NCN = Ala, NAN = Asp and/or Glu, and NGN = Gly. (iv) Later amino acids were added sequentially to the code by a process of subdivision of codon blocks in which a subset of the codons assigned to an early amino acid were reassigned to a later amino acid. (v) Later amino acids were added into positions formerly occupied by amino acids with similar properties because this can occur with minimal disruption to the proteins already encoded by the earlier code. As a result, the properties of the amino acids in the final code retain a four-column pattern that is a relic of the earliest stages of code evolution.

conclusionThe driving force during this process is not the minimization of translational error, but positive selection for the increased diversity and functionality of the proteins that can be made with a larger amino acid alphabet. Nevertheless, the code that results is one in which translational error is minimized. We define a cost function with which we can compare the fitness of codes with varying numbers of amino acids, and a barrier function, which measures the change in cost immediately after addition of a new amino acid. We show that the barrier is positive if an amino acid is added into a column with dissimilar properties, but negative if an amino acid is added into a column with similar physical properties. Thus, natural selection favours the assignment of amino acids to the positions that they occupy in the final code.