A model is presented in which the distribution of low-usage codons in a message is a major factor in determining the impact that they will have on the translation rate and distribution of ribosomes on that message. This model is based on the assumption that low-usage codons are translated more slowly than normal codons, an assumption supported by various lines of published experimental evidence. Although the parameters used to develop this model are somewhat arbitrary, the main conclusions of this paper are consistent with a wide variation in the values of those parameters. In the model, low-usage codons arranged in clusters are much more effective in blocking ribosome movement on the message than ones that are dispersed. The effective size of the cluster is limited to the dimensions of the ribosome. It has been estimated that ribosomes on a message are spaced at least 27 nucleotides or nine codons apart. A ribosome translating a cluster of nine codons in which some or all of the codons are low-usage will move more slowly than over a comparable stretch of message containing no low-usage codons. Owing to ribosome size, the ribosome immediately behind the stalled ribosome will move as slowly; it must wait for the stalled ribosome to move on before it can even begin to translate the difficult region containing the low-usage codons. When the low-usage codon cluster is at the (3^{\prime}) end, the message will eventually be occupied by a ribosome jam that will transmit back to the (5^{\prime}) end of the message. In the steady state, the slowing effect imposed by a cluster of nine low-usage codons at the (3^{\prime}) end of a message would be just as great as if the entire message was composed of them. If the cluster is situated in the middle of a message, the ribosomes will form a jam upstream of the cluster. The ribosome density downstream of the cluster will be considerably reduced from what it would be for the same message with no cluster. If the cluster is at the (5^{\prime}) end of the message, the density of ribosomes will be reduced over the entire length of the message but the overall translation rate per ribosome will be only slightly reduced. However, owing to the reduced number of ribosomes initiating, the efficiency of the message in protein synthesis will be considerably reduced.
The size, number and distribution of low-usage codon clusters has been determined in different species. Between one-quarter and one-half of all reading frames contain at least one cluster with three or more low-usage codons. The wide occurrence of low-usage codon clusters indicates the potential for major impact on translation systems. In Escherichia coli and yeast, clusters containing up to five low-usage codons have been observed. In Drosophila melanogaster and primates, clusters containing up to seven low-usage codons have been observed. Probability calculations indicate a positive correlation between the frequency of low-usage codons and presence of clusters. Detailed calculations for a few naturally occurring messages predict the profound effect of clusters of low-usage codons in these messages on ribosome distribution and translation rates.