When analogies go bad
August 5, 2011 § 5 Comments
Lots has been written about the scientific method (and even I have written about it in a minor way in the past). The cycle of “make hypothesis, make predictions, test predictions, revise hypothesis, repeat” is the main thing people focus on when talking about how scientific progress happens. What’s less talked about is where the hypothesis comes from in the first place, which starts with someone (maybe you, dear reader) noticing something that needs to be explained. This is harder than it may sound, because in order to see something that needs to be explained, you need to be able to see past the existing explanations. You need to notice that what the textbook says should happen isn’t quite correct, instead of falling prey to the temptation to edit what you’re seeing to match what you expected to see. You need, in short, to creatively ignore dogma. And so it’s always interesting to me to watch what happens as a dogma begins to shift. The shift may be of earthquake proportions, as when a new way of looking at a problem causes you to doubt everything you ever knew — what people talk about (but far too often) as a paradigm shift. Or it may be more of an evolution of your understanding — the new idea conflicts with dogma, but there’s no fundamental reason why it should. Your world view can accommodate the new idea without major changes; it’s just that you didn’t think it had to.
Today’s dynamic dogma has to do with the ribosome, and the shift is of the latter type. As you all know (and if you don’t, Wikipedia will tell you) the ribosome is responsible for reading the information carried in messenger RNAs and translating it to produce proteins. It’s a big, complicated protein/RNA complex, and we think of it as both homogeneous and hidebound, a stereotyped machine with a single job: take RNA as input, produce protein as output. We don’t think about the ribosome as exerting any control over which RNAs get translated — or, at least, we haven’t until quite recently. But evidence is accumulating that it can. The most recent chapter in this story is some rather dramatic evidence that mutations in a specific ribosomal subunit can cause substantial changes in the vertebrate body plan (Kondrashov et al., 2011 Ribosome-mediated specificity in Hox mRNA translation and vertebrate tissue patterning. Cell 145 383-397). It looks to me as if we’re going to have to start thinking about the ribosome as an active participant in the regulation of gene expression.
Strictly speaking, the notion that ribosomes aren’t always the same, and don’t always do the same thing, has been around for a while. In the slime mold Dictyostelium discoideum, known to its friends as Dicty, ribosomes were shown to have different physical properties at different stages of the organism’s life cycle. We know that the ribosomal RNA can be modified, as can the proteins; modifications of ribosomal RNA were shown last year to alter how efficiently the ribosome uses an internal ribosome entry site that helps to control the synthesis of the tumor suppressor p53. So, a mutation in a gene involved in modifying ribosomal RNA can make you more susceptible to cancer. Mutations in specific ribosomal protein genes can lead to developmental defects in both plants and zebrafish. One of the clearest stories about a change in ribosomal leading to a change in ribosomal function came from the Silver lab, who showed that some of the duplicated ribosomal genes in yeast have functional differences between the two copies. A subset of the duplicated genes are required for the localization and translation of an mRNA involved in bud site selection. Admittedly, in many of these situations it was hard to be sure that the effect on translation was actually specific. You could argue, for example, that the effect of a modification of ribosomal RNA is to slow down translation of all proteins; if p53 seems to be specifically affected, that’s just because the expression of p53 is unusually sensitive to the general slowdown. And so the idea that ribosomes can be heterogeneous, and that ribosomes with different compositions may behave differently, is not one that has general currency.
The story of the new evidence starts with a mutant mouse phenotype, called Tail Short. Heterozygous mice carrying this mutation have short, kinked tails and various skeletal mutations, including the presence of an abnormal 14th rib. Homozygous embryos die very early. Kondrashov et al. used positional cloning to trace the mutation responsible to the gene encoding a ribosomal protein, Rpl38. The obvious explanation for a defect like this is that it results from a change in the expression of the Hox genes that are the key regulators of skeletal patterning. Expression changes could result either from changes in transcription, or changes in translation. The possibility that transcription was affected by this mutation was easy to rule out: mRNAs from the Hox genes were expressed in the right places, at the right times, in the mutant mice. The tools to study changes in translation in a developing embryo are not so advanced, however. The authors wanted to ask two questions: (1) is the translation of Hox mRNAs affected by the mutation in Rpl38? and (2) if so, is this a side effect of a change in the global level of translation, or is it specific?
To ask whether global translation is affected in these mice, Kondrashov et al. developed a “translation reporter mouse” that carries a gene construct designed to report separately on the two main mechanisms of translation initiation: translation initiated at the mRNA cap, and translation initiated at an internal ribosome entry site. They backcrossed these mice with Tail Short mice, and found that the Rpl38 mutation made no difference: the global level of mRNA translation was unaffected. To look at the translation of specific Hox genes, they then dissected out sections of the developing embryo and used polysome analysis to determine how actively the mRNAs represented in the tissue were being translated. Here, they saw a difference: in the mice carrying the Rpl38 mutation, the number of ribosomes carried on the mRNAs expressed from several Hox genes is significantly reduced, and thus the rate of translation of these genes is reduced as well.
Not only is the expression defect restricted to a subset of Hox genes, it’s also restricted to this particular ribosomal protein: mutations in other ribosomal proteins showed no difference in Hox gene expression. An unexpected finding is that Rpl38 is variably expressed, and is enriched in places where Hox gene expression needs to be high. Other ribosomal proteins are differentially expressed too, though in different patterns. Perhaps they’re regulating the translation of genes other than the Hox genes.
The only hint at mechanism Kondrashov et al. offer is some evidence that Rpl38 is important in assembling the 80S ribosomal particle on Hox mRNAs. (I am not complaining, please note: there is a huge amount of work in this paper already). It’s not clear why these particular mRNAs need Rpl38 for efficient ribosome assembly. Maybe there is a specific sequence or secondary structure on the mRNA that interacts with Rpl38; more to come, no doubt.
What is most interesting to me about all this is, why did we ever imagine that the ribosome was such a passive participant in translation in the first place? This is an old, old piece of biology and it has accumulated many potential variants over the millennia. Why did we imagine that in this case the variants wouldn’t be functionally important? In every other place you look, evolution has exploited variation to add another level of control (at least, in some circumstances). Why should the ribosome be different?
Personally I think that part of the explanation for why we haven’t seen this before lies in the misleading meme of the molecular machine. Bruce Alberts was the first to use the analogy of the “molecular machine” to describe the highly coordinated behavior of protein complexes such as the DNA replication machinery or the ribosome. The ribosome is not just a collection of individual proteins, it’s a single functional entity (to a first approximation) that happens to have been constructed from a number of distinct pieces. Hence the machine analogy. It’s a powerful metaphor that expresses something important about the way biology works, and that’s why we all use it. But it does tend to make you think about a solid, purpose-built machine with crisp edges, like a car engine, not the dynamic, fuzzy, multipurpose “machines” of biology, whose activities are changeable and dependent on context. The more powerful a metaphor is, the more likely you are to find yourself unable to see beyond it.
One of the challenges of this post-genomic era of biological research is that we seem to be running out of helpful analogies. Engineering metaphors (e.g. molecular machines) have taken us a long way, but are running out of steam. The new fashion is to draw analogies from control theory, or computer science; these are potentially useful too, and will probably lead to new dogma being laid down, but will sooner or later turn out to be inaccurate. (The cell is not a computer, either.) That’s the point at which the dogma will need to be creatively ignored in order to make progress.
Kondrashov N, Pusic A, Stumpf CR, Shimizu K, Hsieh AC, Xue S, Ishijima J, Shiroishi T, & Barna M (2011). Ribosome-mediated specificity in Hox mRNA translation and vertebrate tissue patterning. Cell, 145 (3), 383-97 PMID: 21529712