The study started not with experiments but theory. “The ‘aha!’ moment was lying in bed,” says Drummond. For the theory to work, however, the threat of errors would have to be large enough. Drummond checked his references for translation error rates. “They were absurdly high,” he says.
Ribosomes make ∼5 errors per 10,000 codons translated. “When you convert that into how many proteins are translated,” says Drummond, “it becomes much more interesting to look at the cost.” With this error rate, ∼19% of average-length yeast proteins should have a missense error; conservatively, 5% of the proteins might misfold. For the abundant PMA1 transporter that raises the scary prospect of ∼63,000 misfolded molecules per cell.
The team found that misfolding matters not because it takes proteins out of circulation, but because it creates a potentially damaging molecule. What mattered most for evolution rate was not absolute abundance (the more abundant, the more activity that can potentially be lost by misfolding) but translation frequency (the more translation events, the more opportunities for creating troublesome and long-lived misfolded proteins).
The theory also makes sense of proteins such as the plant enzyme Rubisco, perhaps the most abundant protein on Earth. It is highly conserved but the result of this slow evolution is not functional fragility but, the authors hypothesize, translational robustness—surprisingly few inactivating mutations have been found in its gene.