Joined: May 2002
One more article.
Here's the short version of the case as I now understand it.
- In the beginning, scientists thought the genetic code was universal (maybe; this is the standard line, whether all relevant experts also assumed this initially seems to me to be uncertain, at least I've not seen any analysis of the topic).
- in the 1980's it was documented that this was not the case
- In the late 1980's Osawa proposed the "codon disappearence" theory for the evolution of code changes, described in the Schultz & Yarus (1996) article referenced below thusly:
Codon reassignment to new amino acids within large, complex, relatively modern genomes (Osawa et al. 1992) poses interesting mechanistic problems. Osawa and Jukes have proposed (1989), and reaffirm in recent publications (1992; Osawa and Jukes 1995), that during codon reassignment every example of a codon in an entire genome must mutate or otherwise disappear as a result of mutational change in genomic GC content. Subsequent to its total disappearance, a codon can be captured by, e.g., an anticodon mutation in a dispensable tRNA, thereby reappearing with a new identity. We will call this the ``codon disappearance'' theory, after its characteristic intermediate state.
- In the mid-1990's another theory was proposed, apparently right in Schultz & Yarus' 1996 article:
We find the absolute disappearance of hundreds, thousands, or tens of thousands of examples of a codon by mutation pressure alone, in diverse independent cases, an improbable evolutionary scenario. Total disappearance should be an extremely slow occurrence, because mutation pressure and genetic drift in large populations are among the weakest evolutionary forces, producing only very slow changes in genomic composition. Furthermore, back mutation increases in effectiveness as the goal is approached because of the accumulation of codons related to the disappearing codon by single mutation. Finally, complete disappearance of codons in eukaryotes would be hindered by coherent areas of varied GC content along chromosomes (Sharp and Lloyd 1993; Ikemura and Wada 1991). Because codon choice follows GC content, such areas can provide sheltered enclaves for particular codons (Santos and Tuite 1995).
Though total disappearance is difficult to prove, mutation pressure certainly causes codon frequencies to change. Evolution to very low frequencies and inefficient translational function is well supported (e.g., Kano et al. 1993). But we argue that mutation and drift in codon frequency over entire genomes are vulnerable to being overtaken by faster evolutionary processes such as selection. Thus the question: Are there plausible faster processes, perhaps selection-driven processes, for codon reassignment?
Schultz and Yarus characterized a nonanticodon tRNA site (1994a,b) where particular nucleotide sequences allow a tRNA to read an unusual near-cognate codon. More generally, several sites are known where single mutations in nonanticodon nucleotides (reviewed in Yarus and Smith 1995) enhance tRNA ability to read (at least) two codons, (at least) one of which is forbidden by normal base-pairing and wobble rules. Schultz and Yarus suggested (1994c) that such equivocal adapters could catalyze codon reassignment for one of the codons being ambiguously read. (For clarity in what follows, a codon read in more than one way is said to be ``ambiguous''; a tRNA which reads normal codons as well as codons not normally assigned is said to be ``equivocal.'' In the particular case in which reassignment is initiated by a mutation that impairs normal translation of a codon, reassignment via an equivocal adapter tRNA might evolve quickly by selection for improved translation of the newly ambiguous codon. Transitional coding ambiguity could finally be resolved by, for example, loss or mutation of the original tRNA, and anticodon mutation to equivocal complementarity in the new (equivocal) tRNA, so that the amino acid of the previously equivocal tRNA is reassigned. We will call this the ``ambiguous intermediate'' theory.
Here is the reference, and some of Schultz & Yarus' (1996) lines of evidence for the ambiguous intermediate theory:
J Mol Evol 1996 May;42(5):597-601
On malleability in the genetic code.
Schultz DW, Yarus M.
To explain now-numerous cases of codon reassignment (departure from the "universal" code), we suggest a pathway in which the transformed codon is temporarily ambiguous. All the unusual tRNA activities required have been demonstrated. In addition, the repetitive use of certain reassignments, the phylogenetic distribution of reassignments, and the properties of present-day reassinged tRNAs are each consistent with evolution of the code via an ambiguous translational intermediate.
Firstly: at the heart of our proposal lies the supposition that codons are read ambiguously by two tRNAs (or a tRNA and an RF, in the case of terminators), specifying insertion of more than one amino acid (or an amino acid as well as stop). In contrast, the assumption that codons vanish before reassignment, which is characteristic of codon disappearance theory, is mandated by the assertion that codons cannot have two meanings.
In strict form, this axiom of nonambiguity contradicts chemical principle. An infinite free energy difference between reaction pathways is required to select one reactant and reject another absolutely. The strict absence of ambiguity is also contradicted by experiment. Cumulative missense translation in normal E. coli has been estimated at 4 × 10-4 per codon (Ellis and Gallant 1982). Total miscoding per peptide chain is the much larger sum over the hundreds of codons in the protein. Therefore an appreciable basal ambiguity (yielding ~ 10% of the average 250-amino-acid protein with a variant sequence) is evident, and tolerated, in wild-type cells.
Further, cells are unharmed even when this substantial basal ambiguity is increased dramatically. We have constructed strains containing equivocal E. coli tRNAs that demonstrate suppressor efficiencies of 50 to nearly 100%, making a stop codon ambiguous (Schultz and Yarus 1994a,b). Ribosomal ambiguity mutations (RAM) increase misreading of stop codons up to 100-fold in cells that remain viable (Strigini and Brickman 1973; Andersson et al. 1982). Most specifically, the general error frequency can be increased 13-fold (using 5 µg/ml streptomycin) and cells continue to grow exponentially at a rate close to controls. After more than 400 generations in streptomycin, there is no detectable decrease in cellular viability (Gallant and Palmer 1979). Thus ambiguity at a variety of codons (to >=1 error in the average 250-amino-acid protein) is well tolerated, or has no apparent phenotype. The limited ambiguity we posit as the initiating event in codon reassignment, occurring at one (or a few) codon(s) and perhaps initially quantitatively small, seems quite plausible in this context.
Nor is coding ambiguity limited to prokaryotes. Eukaryotes have basal levels of coding ambiguity which are probably similar to prokaryotes (Gallant and Palmer 1979). Normal yeast glutamine tRNAs are known to read equivocally at the first codon position (Weiss and Friedberg 1986; Edelman and Culbertson 1991). Similar ambiguities can be exploited for an organism's own purposes, as when animal and plant viruses purposefully use ambiguous stop condons to adjust the level of stop readthrough to an essential gene product. This misreading by a wild-type tRNA is known to approach 5% at stop codons within a special mRNA context (Skuzeski et al. 1991; Feng et al. 1990). Thus, during codon reassignment there seems to be no reason that all codons must invariably be read without ambiguity.
[note here that ambiguity is not exactly vanishingly rare and therefore the assumption that intermediates would be nonviable is false]
Secondly: There is no definite direction to reassignment in codon disappearance theory; dispensable RNA genes may capture unassigned codon by, e.g., random single base changes in their anticodons (Osawa and Jukes, 1989). However, we first argue that known reassignments (Table 1) are very nonrandom. We then argue the nonrandomness supports ambiguous intermediate theory because it is explicable by types of equivocal reading already demonstrated in tRNAs.
However, 14 of 14 single-nucleotide reassignments in Table 1 parallel the activities of known equivocal tRNAs. That is, all 14 changes might be mediated by tRNAs reading a single base equivocally, using G-U (anticodon-codon) wobble at the first position, or C-A or G-A mispairing at the third codon position. Equivocal C-A third-position mispairing has long been known from study of tRNA opal (UGA) suppressors (Hirsh 1971). We recently constructed two new tRNAs that demonstrate in vivo the required equivocal G-U and C-A readings (Schultz and Yarus 1994b), thereby potentially accounting for ten assignments (Table 1). This congruence, in fact, first drew our attention to the possibility that tRNAs might mediate codon reassignment. The remaining exceptional wobble, transitional G-A pairing at the third position, has also been detected in the equivocal tRNA repertoire in vitro, using cytoplasmic and chloroplast tRNA Cys (Nicotiana) as UGA suppressors (Urban & Beier, 1995). The remaining 15th case requires a more unusual first/second-position double miscoding. However, the Candida albicans tRNA translating the reassigned CUG codon has been independently shown to be capable of a similar doubly equivocal coding (Santos et al. 1993; see below). Thus 15 of 15 known reassignments can be matched with known tRNA capabilities.
Thirdly: Phylogenetic distribution of reassignment is consistent with ambiguous intermediates. Tourancheau et al. (1995) have made the initially surprising observation that UAA/UAG in ciliates have been reassigned to glutamine at least three times independently (on the basis of the rRNA tree), instead of depending on a common ancestral reassignment. This striking phylogenetic cluster of identical but independent reassignments has no apparent explanation in the codon disappearance scheme. However, such a cluster is easily explained within the ambiguous intermediate mechanism by a tendency to equivocal reading of these codons inherited from an ancestor. Such an ambiguity might be conserved within a group of species if used for an important regulatory event like stop codon readthrough. These authors also found no correlation between GC content of the ciliates and reassignment, which might have been expected if evolutionary change in GC content drives the process.
Fourthly: Molecular fossil and functional evidence of translational ambiguity accompanies known cases of reassignment. We have previously pointed out that sequenced tRNAs that have captured new codons, such as the UAA and UAG reading tRNAs from the ciliate Tetrahymena thermophila (Hanyu et al. 1986), contain unusual nucleotide sequences that we have identified as enhancers of equivocal coding in E. coli (Schultz and Yarus 1994c). Thus the structure of these three related isoaccepting tRNAGln sequences suggests the existence of an ancestor that coded equivocally.
In summary: We acknowledge the significance of codon reassignment, and do not argue against change in GC content as a significant evolutionary event (e.g., Sueoka, 1993). But we do argue that codon reassignment is unlikely to be carried out entirely by the slow processes of mutation pressure and drift. Additionally, the axiom of nonambiguity fundamental to codon disappearance theory is not justified. The evident nonrandomness of known reassignments, the clustering of similar changes in phylogeny, and the properties of reassigned tRNAs, where known, are strikingly consistent with ambiguously translating intermediates. These phenomena are unexpected or contradictory to codon disappearance theory, acting in isolation.
In this connection, there is no logical incompatibility between mutational change in GC content and ambiguous intermediates. Schultz and Yarus (1994c) have noted that these may occur together. In fact, a codon which has become rare might also be expected to evolve a rare cognate tRNA. Such a rare tRNA would be more vulnerable than usual to competition during translation, including competition from an equivocal adaptor translating its codon. Thus not only might mutation pressure be overtaken by faster selection, but the initial effects of mutation pressure might facilitate the overtaking mechanism. Quantitative modeling of this process might prove rewarding.
Finally: if ambiguous intermediate theory gives a good account of modern coding changes, it thereby becomes a preferred route by which a limited ancestral code could have been transformed to the present ``universal'' genetic code. In fact, coding transitions via ambiguous intermediates would likely be easier during the formation of the code than today.
Other aspects of ambiguous intermediate and codon disappearance schemes can be compared in the previous note by Osawa and Jukes (1995), and in Schultz and Yarus (1994c), to which the interested reader is directed for references and details which do not appear here.
Edited by niiicholas on May 30 2002,01:18