Not every polymorphism close to the AUG codon can be explained by invoking context effects on initiation of translation

The −1C>T change in the annexin V gene, which reduces the risk of myocardial infarction,1 cannot be explained as simply as your editorial note suggests.2 

The demonstration by Gonzalez-Conejero et al1 that −1T increases the efficiency of translation might be correct (the fourth paragraph of this letter discusses some concerns about the in vitro translation assay), but if so, the effect would not be explicable by invoking the Kozak context rules. The rules predict the opposite of what was seen; that is, translation should be more efficient with C rather than T in position −1, if that position scores at all.

GCCRCCaugG (R = purine) is the optimal context for recognition of the start codon in mammals. Within this motif, some positions are more important than others. Mutagenesis experiments with laboratory constructs showed that A or G in position −3 (3 nt upstream from the AUG codon, which is numbered +1 to +3) and G in position +4 make the strongest contributions.3,4 Only in the absence of −3R and +4G do mutations in positions −1 and −2 score strongly.3 Inasmuch as the annexin V start site conforms to the consensus motif in positions −3 and +4, the identity of the base in position −1 would be expected to affect translation only slightly, if at all.

The context rules, initially established by studying translation in cultured cells,3,5 can be replicated using in vitro translation systems, but the experiments must be designed carefully. Reaction conditions, such as the concentration of magnesium,4,6 can profoundly affect whether recognition of the AUG codon in vitro displays the same sensitivity to context as is seen in vivo. The coupled transcription/translation system used by many investigators, including Gonzalez-Conejero et al,1 makes it difficult to adjust magnesium levels and to be sure that exactly the same amount of mRNA is generated from each allele. A small change in translational efficiency is believable when the reduction in initiation from the first AUG codon is accompanied by an increase in initiation from the next AUG downstream (eg, Figure 1 in Kozak7) or when an assay is used that directly monitors the initiation step.4 One cannot be as confident, however, about a small (1.4-fold) difference in translational efficiency based on measurement of proteins precipitable by trichloroacetic acid.1 

In addition to these practical complications in testing for context effects, there are theoretical limitations. Because secondary structure downstream from the AUG codon can compensate for a less than perfect context,8 not every mutation within the consensus motif will score.

In some human and mouse genes, a mutation or polymorphism close to the AUG codon, usually in position −3 or +4, has been shown to reduce translational efficiency, with pathological consequences.9 Only one previous example involves a change in position −1: in a patient with ataxia with vitamin E deficiency, a C>T mutation in the alpha-tocopherol transfer protein gene causes a 2-fold decrease in translation, measured in vivo.10 In addition to these pathologies linked to mutations in the consensus motif, there is a growing list of human diseases wherein translation of a critical regulatory gene is perturbed by restructuring the 5′ UTR in ways that add or remove upstream AUG codons.9,11 Thus, the scanning mechanism for initiation of translation does provide a framework for understanding how some mutations cause disease.

It is possible that the demonstrated increase in plasma levels of annexin V protein associated with the −1T allele1reflects an effect on mRNA stability or splicing. If followup studies rule out these alternative explanations and confirm that the −1T allele indeed augments translation, the reason could conceivably involve an effect on mRNA secondary structure. I do not think this particular polymorphism can be explained by invoking conventional context effects on AUG codon recognition.