Mistakes in translation don't translate into termination

The evolution of the genetic code remains one of the greatest mysteries in biology. Since the elucidation of the code in the 1960s many hypotheses have been generated to try to explain the assignment of the 64 codons to the canonical 20 amino acids and punctuation. Perhaps the most famous of these, posited by Francis Crick (1), is the “frozen accident” hypothesis, in which the associations of amino acids with their three base codons evolved haphazardly, were fixed in place as organisms became more complex, and thereafter could change only with great difficulty. In this scenario, many mutations would result in amino acid substitutions that would greatly impair the functionality of proteins. Alternatively, the genetic code may have undergone a period of optimization before fixation, and amino acid substitutions would be more chemically and functionally conservative.

Choosing between these (and other) scenarios is extremely difficult, though, because all of biology has evolved for many billions of years in the context of the almost universal code and thus has already been highly optimized for the extant code irrespective of whether there was preoptimization or not. At best, we can examine the nature of the chemical constraints on the current genetic code via perturbation and directed evolution experiments. Although many attempts have been made to alter the genetic code by amino acid replacement (2, 3) and codon reassignment (4, 5), few have looked at the global effects of altering the genetic code on an organism until now. To probe the degree and nature of selective pressures that constrain the genetic code, Bacher et al. (6) in this issue of PNAS explore the effects of an editing-deficient isoleucine tRNA synthetase on the growth of Escherichia coli.

To help maintain amino acid specificity, many aminoacyl tRNA synthetases contain a secondary active site that is responsible for editing mischarged or noncognate amino acids before they are used in translation (7-9). Disabling editing sites in these synthetases results in increased nonspecific charging of cognate tRNAs with amino acids that are sterically similar to the cognate amino acid. In particular, Decrècy-Lagard and coworkers (10) have previously knocked out the editing function of the isoleucine tRNA synthetase and shown that isoleucyl tRNA is charged equally well with either isoleucine or valine. The editing-deficient strain essentially creates an “ambiguous” genetic code that generates numerous missense proteins. Bacher et al. (6) now report, not surprisingly, that the abolition of the editing function in the isoleucine tRNA synthetase decreases the growth rate of E. coli over a range of growth temperatures and media conditions.

Although the folding catastrophe is widespread, it is also relatively benign.

However, it is the nature of the growth defects that are most interesting in terms of constraints on the genetic code. To the extent that the code is a frozen accident, chemical changes might be expected to lead to the biochemical equivalent of a genetic “error catastrophe,” in which numerous proteins are misfolded or unfolded. Supporting this hypothesis, the editing-deficient strain shows increased sensitivity to antibiotics that target the ribosomal A site, cell wall biosynthesis, and DNA replication (6), indicating that codon ambiguity has widespread effects on protein function. The global effects on fitness that are observed after editing failure are consonant with the phenotypes observed after growth on amino acid analogues. For example, replacement of tryptophan with tryptophan analogues in both E. coli (3) and the bacteria phage Qβ (11) resulted in decreased growth. The growth defects could only be partially compensated for by multiple mutations in multiple proteins, perhaps improving their ability to fold in the presence of the analogues. In contrast, though, Döring and Marlière (12) have reported that cysteine can replace isoleucine in E. coli with only a minimal effect on growth rate. Cysteine's relatively small size may allow it to better substitute for isoleucine at sterically constrained sites in proteins than does valine. The varied outcomes of these various growth experiments suggest that different amino acids and even different positions in the genetic code will be differentially sensitive to substitution. Further exploration of these differences may provide insights into the evolution of alternative codes, such as those that exist in mitochondria.

What is surprising, though, is that, although the folding catastrophe is widespread, it is also relatively benign, at least in terms of the magnitude of its effects on cell biology. Cells grow more slowly, but they do grow, even with half of their isoleucine residues being miscoded. In addition, it might be supposed that increasing the number of proteins that are improperly folded might severely impact the ability of a cell to replicate not only efficiently, but with fidelity. However, the mutation rate in the editing-deficient mutant did not increase significantly when compared with WT. Moreover, the editing defects did not cause the ribosome to introduce more frameshift errors than normal. Taken together, these results may argue that the genetic code is so supremely optimized that the effects of any folding catastrophe are greatly reduced.

Moreover, there is increasing evidence that not only is systemic misfolding less of a problem than might be imagined, but organisms have evolved mechanisms to confront and even take advantage of systemic misfolding. The heat shock response is one well known example, and protein chaperones, such as Hsp90, act to refold various misfolded proteins. In this regard, it would be interesting to see whether organisms challenged with unnatural amino acids induce the same types of expression changes as organisms challenged with heat. But beyond the simple repair functions embodied by chaperones, even more dynamic responses to misfolding may be present in cells. For example, Saccharomyces cerevisiae has evolved a novel system to actually exploit hidden genetic variation in the organism via a conformational mutant of the translation termination factor Sup35p [PSI⁺] (13). A yeast strain that contains the [PSI⁺] element possesses a short-term survival advantage over strains that lack it, as stress induces the alternative conformation of [PSI⁺], leading to read-through of termination codons and the appearance of alternate and potentially beneficial phenotypes (14). To the extent that eukaryotes have, to varying degrees, evolved the ability to exploit translational defects and protein misfolding, might prokaryotes also have such mechanisms? Do conditions that lead to protein misfolding in E. coli (such as those reported by Bacher et al.) actually offer any benefit to organismal survival?

Taking advantage of protein misfolding might at first seem to be an improbable event, but this phenomenon is conceptually similar to other ways in which organisms take evolutionary advantage of even inclement environments. For example, starvation can lead to increased mutation rates, the so-called adaptive mutation phenomena, by a variety of mechanisms (15, 16). It can thus be argued that, when organisms are placed under environmental stress, there is a general need to explore a larger genetic space or a larger protein folding space or both. To the extent that organisms have encountered environmental stress intermittently over evolutionary time, it may even be advantageous to establish some sort of regulatory feedback between stress and phenotypic exploration. As examples, some types of hypermutation can occur only upon the stress-induced production of mutator polymerases, and some Candida species appear to enhance their resistance to environmental stress, by ambiguously substituting serine for leucine (17, 18).

Nonetheless, although organisms may be resilient to short-term recoding of amino acid identity, Bacher et al.'s results confirm that, over the long run, there has been and will continue to be tremendous selective pressure to maintain the current genetic code. Recent attempts to alter the genetic code have wisely focused on recoding amber termination codons (UAG) to insert natural or unnatural amino acids (4, 5). Going beyond these few, simple substitutions will be an uphill battle. But with the example of mitochondrial codes to spur us on, and the assurance of Bacher et al. that faithless incorporation slows but does not extinguish life, it should be possible to use directed evolution to completely recast the genetic code.