My friend Mike Hecht at Princeton has just published a quite interesting paper in PLoS One (also known as the journal that takes significant findings when no one else will). In this paper (Fisher et al. (2011), 6:e15364) he shows that an expression library of completely random ‘proteins’ of 102 residues in length can suppress a variety of deletion variants that would otherwise inhibit cell growth. Most remarkably, four different deletion variants can be suppressed by the tandem expression of four selected, but otherwise random, proteins.
Now, there are various ways to view these results, as Hecht and colleagues indicate. One possibility is that the selected proteins replace the activities of the deleted proteins (i.e., classic suppression). This is the most interesting and also the most implausible result. It would be truly remarkable if 102 amino acid proteins drawn from a relatively small library of 10^6 different four helix bundle-like proteins had even a fraction of the catalytic activity of biosynthetic enzymes groomed by eons of evolution. My incredulity stems from the difficulties that protein engineers have long had with generating enzymatic activities from scratch (recent successes by Baker aside). In particular, it beggars my imagination to believe that citrate synthase, an enzyme that brings together two substrates and catalyzes the formation of a new carbon-carbon bond, can be replaced by muck of length 102. My beggared imagination is supported somewhat by the fact that the when the small, selected proteins are isolated they do not seem to have the missing enzymatic activity.
So, if it’s not that, then what is it? The manuscript goes to some pains to suggest that it’s also not the activation of an alternative pathway for synthesis, and that it’s not the inadvertent activation of another enzyme that can catalyze the same reaction. Matsumura and his group at Emory carried out a really neat study in which they looked at how some enzymes when expressed at high levels could suppress deletions of other enzymes; that is, they determine whether enzymes could ‘moonlight’ for one another (Patrick et al. (2007), MolBiolEvol 24:2716). They found many examples, but most of these made some sense, in that the enzymes identified had basal activities that were similar to the deleted enzymes. For example, some phosphatases were unsurprisingly found to have loose enough substrate specificities that they could fill in for other phosphatases.
In the end, they don’t know what the mechanism is, which makes many of us (myself included) rather queasy. I continue to believe that enzymes are not easy to make, and yet the results speak for themselves: activity comes from somewhere. There are various other mechanisms that may be plausible, but they sort of boil down to this: complex systems can do complex things. That is, when there are many enzymes around, a new protein may be able to alter another protein’s substrate specificity, induce expression of another enzyme (although, again, not the ones that would have been expected based on what Patrick et al. found), or bring together various other proteins to form new interactions that may fortuitously catalyze new reactions.
So, let’s just go with that: complex systems have many, unanticipated, untapped phenotypic states, and these phenotypic states can be accessed by something as simple as the expression of a random protein. If this is true for something as basic as biosynthetic pathways, might it also be true for other aspects of metabolism, such as pathogenesis or immune escape? Is it possible to change the state of what would otherwise be a non-pathogen to become a pathogen solely by randomly expressing some information? Again, the rational view would be ‘no,’ since presumably many host:pathogen interactions (or pathogen:immune interactions) are predicated on a very delicate and refined dance between pathogen surface markers and host receptors (or immune molecules). And nominally one might think that just throwing in 102 amino acids here or there would not necessarily be the same as crafting a surface glycoprotein that could intimately interact with, say, the transferrin receptor. But really after the Hecht paper, who knows? Maybe there are many otherwise cryptic pathogen markers whose conformations can be altered, whose expression can be induced, or that can be formed into complexes that are much more than the sum of their parts.
In this regard, it’s probably also good to remember that the capacity of the complex cellular genome for expression has also been plumbed, and has again been found to be surprisingly responsive. Several groups have now made ‘artificial transcription factors’ (for example, by randomly fusing zinc finger domains) and then transfecting them into tissue culture cells, just to see what they would do (see, for example, Park et al. (2003), NatureBiotech 21:1208; Blancaford et al. (2003), NatureBiotech 21:269). Perhaps less surprisingly than the Hecht example, these authors found that they could screen for new combinations of DNA-bindiing domains that could turn on and off different genes, and that could elicit novel phenotypes (including fun ones such as drug resistance) in the transfected cells.
So, while we continue to get in a tizzy about the dangers of making long, engineered DNA circuits, here in the background are random screens that are turning up completely unknown and unexpected phenotypes. No one has yet looked for an impact on pathogenesis (so far as I know, but I’m probably just a literature search away from having that ignorance overturned) or immune evasion, but it’s almost certainly possible. Synthetic biology is as nothing to synthetic systems biology, which will rely on bringing out hidden organismal potential, rather than upon jerry-rigging some complex new pathway that may or may not do what it’s designed to do.