Joined: Feb. 2006
I had a chance to read the Cell paper I mentioned a few days back. I thought it was fascinating, and I highly recommend it.
Basically, it appears that a number of things correlate with mRNA levels across species from E. coli, yeast, worms, flies, mice, to humans. These include the fraction of optimal codon usage (i.e., the fraction of codons that correspond to the most abundant tRNA for that amino acid), the evolutionary non-synonymous substitution rate, the synonymous substitution rate, and even the relative rate of transitions to transversions.
The authors use principal component analysis to argue that all of these are related to one main underlying feature. They then argue that this feature is the need to minimize translation errors that lead to protein misfolding. In essence, they argue that misfolded proteins are cytotoxic, presumably in rough proportion to their abundance.
For low abundance proteins, occasional misfolding contributes little to the total cytotoxic burden in the cell. But for high abundance proteins, even rare misfolding may be detrimental. Thus, they argue, highly expressed genes need to use optimal codons to minimize translation errors. Even synonymous substitutions in a highly expressed gene can be detrimental, because they will tend to change an optimal codon to a suboptimal codon. This will increase the rate of translational error, resulting in more misfolded proteins, and greater cytotoxicity.
They go on to show how all of the observed correlations with gene expression level can be explained by this underlying mechanism. They do simulated evolution studies in silico that reproduce the observed correlations, but only if they include a cost associated with protein misfolding. Then they go a step further and suggest that this effect shows tissue specific features in complex organisms. For example, they suggest that neural tissue may be particularly sensitive to cytotoxicity from misfolded proteins (think Alzheimer's, Parkinsons, CJD, etc.). They note that brain-specific genes appear to evolve relatively slowly as a group, and explain how their hypothesis accounts for this.
The authors suggest that, if they're right, this has wide ranging implications for our understanding of evolution. Among other things, we would need to take into account how this affects synonymous vs. non-synonymous substituion rates when estimating divergence based on molecular data.
Here's what I really liked about this paper. 1) It proposes a new mechanism that has fundamental implications for how evolution works and is constrained. (At least, it's new to me; an editorial in the same issue of Cell seems to think it's potentially quite important as well.) 2) It provides a unifying explanation for a number of seemingly unconnected observations. (It even provides possible insight into the mechanisms of type 2 diabetes!) 3) The authors make multiple predictions based on their proposal, all of which can be tested experimentally.
To me, this is a stellar example of how science really works. The contrast with ID is stark.