The universality of the genetic code is striking, but some exceptions exist. They tend to affect the codons involved in initiation or termination and result from the production (or absence) of tRNAs representing certain codons. The changes found in principal (bacterial or nuclear) genomes are summarized in Figure 7.10.
Almost all of the changes that allow a codon to represent an amino acid affect termination codons:
Acquisition of a coding function by a termination codon requires two types of change: a tRNA must be mutated so as to recognize the codon; and the class 1 release factor must be mutated so that it does not terminate at this codon.
The other common type of change is loss of the tRNA that responds to a codon, so that the codon no longer specifies any amino acid. What happens at such a codon will depend on whether the termination factor evolves to recognize it.
All of these changes are sporadic, which is to say that they appear to have occurred independently in specific lines of evolution. They may be concentrated on termination codons, because these changes do not involve substitution of one amino acid for another. Once the genetic code was established, early in evolution, any general change in the meaning of a codon would cause a substitution in all the proteins that contain that amino acid. It seems likely that the change would be deleterious in at least some of these proteins, with the result that it would be strongly selected against. The divergent uses of the termination codons could represent their "capture" for normal coding purposes. If some termination codons were used only rarely, they could be recruited to coding purposes by changes that allowed tRNAs to recognize them.
Exceptions to the universal genetic code also occur in the mitochondria from several species. Figure 7.11 constructs a phylogeny for the changes. It suggests that there was a universal code that was changed at various points in mitochondrial evolution. The earliest change was the employment of UGA to code for tryptophan, which is common to all (non-plant) mitochondria (for review see Osawa et al., 1992).
Some of these changes make the code simpler, by replacing two codons that had different meanings with a pair that has a single meaning. Pairs treated like this include UGG and UGA (both Trp instead of one Trp and one termination) and AUG and AUA (both Met instead of one Met and the other Ile).
Why have changes been able to evolve in the mitochondrial code? Because the mitochondrion synthesizes only a small number of proteins (~10), the problem of disruption by changes in meaning is much less severe. Probably the codons that are altered were not used extensively in locations where amino acid substitutions would have been deleterious. The variety of changes found in mitochondria of different species suggests that they have evolved separately, and not by common descent from an ancestral mitochondrial code.
According to the wobble hypothesis, a minimum of 31 tRNAs (excluding the initiator) are required to recognize all 61 codons (at least 2 tRNAs are required for each codon family and 1 tRNA is needed per codon pair or single codon). But an unusual situation exists in (at least) mammalian mitochondria in which there are only 22 different tRNAs. How does this limited set of tRNAs accommodate all the codons?
The critical feature lies in a simplification of codon-anticodon pairing, in which one tRNA recognizes all four members of a codon family. This reduces to 23 the minimum number of tRNAs required to respond to all usual codons. The use of AGAG for termination reduces the requirement by one further tRNA, to 22.
In all eight codon families, the sequence of the tRNA contains an unmodified U at the first position of the anticodon. The remaining codons are grouped into pairs in which all the codons ending in pyrimidines are read by G in the anticodon, and all the codons ending in purines are read by a modified U in the anticodon, as predicted by the wobble hypothesis. The complication of the single UGG codon is avoided by the change in the code to read UGA with UGG as tryptophan; and in mammals, AUA ceases to represent isoleucine and instead is read with AUG as methionine. This allows all the nonfamily codons to be read as 14 pairs.
The 22 identified tRNA genes therefore code for 14 tRNAs representing pairs, and 8 tRNAs representing families. This leaves the two usual termination codons UAG and UAA unrecognized by tRNA, together with the codon pair AGAG. Similar rules are followed in the mitochondria of fungi (for review see Fox, 1987).