Eve and the tree of knowledge
June 30, 2010 § 3 Comments
Zeba Wunderlich (DePace lab) pointed me to this paper about the even-skipped (eve) stripe 2 enhancer in Drosophila. (Arnosti DN, Barolo S, Levine M, Small S. 1996. The eve stripe 2 enhancer employs multiple modes of transcriptional synergy. Development 122 205-14 PMID: 8565831), which apparently sparked a lively discussion in journal club last week. It’s an interesting paper that seems to indicate something of a transition point for Drosophila work, namely the point at which Drosophilologists got serious about trying to understand not only which DNA sequences in an enhancer are important, but also why this particular combination of DNA sequences, in this order, produces the quantitative output that it does. And the question is still there for the answering. The bottom line for me, after reading this paper, is that if you had any hope that your personal genome sequence would tell you much about protein levels in the cells of your body, you should give up that hope right now. (Except in extreme cases, of course.) This is an extraordinarily well-studied system where all the important inputs are known, and how the input levels vary from cell to cell in the wild type is also known. Even so, a heroic amount of work produced little in the way of principles that could be used to predict how individual gene sequences (say, in individual humans) might differ in expression patterns. And yet, there are hints in this paper that with better quantitation and the ability to detect multiple inputs and outputs at once, the problem may be tractable. In other words, it’s a great paper to read for background on why the DePace lab, and other labs, are doing what they’re doing.
Don’t misunderstand me — your personal genome sequence, when you shell out $1000 to get it, will certainly tell you some important things. Correlations between specific genotypes and certain types of risks, or drug responsiveness, will be increasingly easy to identify, and increasingly important. And polymorphisms in coding sequences should be relatively easy to interpret, for the most part. But before we can make the kinds of quantitative predictions that you might imagine as the ultimate goal of post-genomic medicine — oh look, this person has a polymorphism in one of the activator sites in the enhancer region for Bid, she’s going to have a 30% reduction in Bid expression in her breast tissue, that means that she’s right on the borderline for efficacy of cytotoxic drug X, probably better not to subject her to the nasty side-effects of drug X, let’s try drug Y instead — we’ll need a better understanding of many biological processes, transcriptional control being high on the list.
Why would you want to use Drosophila to get at these questions? Almost everyone knows that Drosophila is one of the great systems for studying development, and specifically the question of how a more-or-less symmetrical fertilized egg produces the patterns that determine the shape and organisation of the mature insect. An early event in Drosophila development is that the embryo lays down a pattern of seven two-part stripes that are perpendicular to the anterior-posterior (head to tail) axis. Each stripe expresses a specific pattern of gene expression, and each eventually becomes a specific part of the body plan of the mature fly; eve is expressed in the anterior half of each stripe. Zeba kindly supplied some gorgeous pictures of the stripes, left, from the Berkeley Drosophila Transcription Network Project. The blue stripes (top and bottom panels) are eve expression, the red in the center (bottom panel) is Kruppel expression, and each green dot represents a single nucleus.
After many decades of effort, the Drosophila community knows a huge amount about what the inputs are to specific genes (the transcription factors that drive the expression of the gene, where they are expressed, and in what quantity) and what the outputs are (where and when the gene is expressed in the developing embryo, on a cell by cell level). In this particular paper, the question is how the section of the eve enhancer that drives expression in stripe 2 achieves a very sharp change in output at the anterior edge of the stripe. So, let’s look first at the inputs. The stripes are produced by the responses of key genes — such as eve — to varying levels of specific transcriptional activators and repressors. In this case the relevant factors are bicoid, hunchback, giant and Kruppel. Bicoid and hunchback are activators; giant and kruppel, repressors. Bicoid and hunchback are expressed in a gradient across the embryo, with the greatest concentrations at the head end. [How? Bicoid mRNA is selectively deposited at the anterior pole of the developing oocyte by nurse cells in the mother, and the bicoid gradient later drives a hunchback gradient]. Kruppel is expressed in a broad stripe in the middle of the embryo (see figure), and giant in a stripe at each end of the embryo. At the anterior edge of stripe 2, giant levels change by about 2-fold from one side of the dividing line to the other, and bicoid, kruppel and hunchback levels change hardly at all, but the eve mRNA levels change by over 10-fold.
The eve stripe 2 enhancer has 6 activator-binding sites (5 for bicoid, one for hunchback) and 6 repressor-binding sites (3 for each repressor). The authors set out to systematically delete them, move them around, and modify their binding affinities, as well as altering the inputs by adding in fusion proteins, for example bicoid fused to a strong activator (GCN4). One cherished hypothesis that was apparently killed by the results of these experiments was that giant might chiefly act by competing with bicoid for two binding sites where a bicoid site and a giant site overlap. Sadly, it turns out that these two giant sites are much less important to the position of the stripe boundary than the third — which doesn’t overlap with anything. At the time of this paper’s publication, the mechanism by which this third site achieved repression was not known; perhaps it is now. But the bigger take-aways from this set of experiments were (1) the position of each bicoid binding site is not qualitatively important, though it can be quantitatively important; (2) the number of activators bound to the enhancer is more important than what the activators are (6 bicoid sites can work just as well as 5 bicoid, one hunchback — at least at this level of resolution; and artificial activators work too); (3) the affinity of each activator site is carefully tuned by evolution to give an overall level of activation that is not too high for a small amount of giant repression to overcome. Their overall model for how the sharp anterior edge is achieved is that activator binding is “synergistic” in two ways: two activators together bind better to the DNA, and two activators together do a better job of activating the transcription complex. Giant interferes with both levels of synergy, so a small change in giant level will produce a large change in transcription.
I put the scare quotes around “synergistic” in the paragraph above because there is no way the authors can really know whether they’re looking at synergy here. This is a beautiful paper, very clearly presented, that represents a huge amount of work. Nevertheless, it’s ultimately frustrating. Just how large is the difference between the wild-type stripe and the stripe that results when you knock out one of the giant sites? When you add GCN4 to bicoid, how much does that shift the position of giant expression? Is there really synergy between the activators, and if so how much? The whole situation is crying out for better quantitation.
I guess that’s why the DePace lab needs that extremely powerful confocal microscope…
This paper and the careful work of others, such as the DePace lab, may be penetrating some truly life-like properties of biological systems–their power lies not in their optimization but in the way many different paths can lead to a product that is good enough for evolution. Transcriptional regulation with its linear encoding of information and its cybernetic quality, so emphasized by Jacob and Monod, seemed to be the most favorable and the most important process to be converge on a simple physical optimization. The expectation of that optimum permeates physics. It would be disconcerting if a cannon ball could follow a parabolic trajectory some days and a circular one other days to the same effect, or if Snell’s law of refraction was true only for some materials or at certain times. It would be surprising to physicists if it was reported that with this unappreciated flexibility artillery pieces and telescopes would still work pretty well. To understand how transcriptional circuits work and how they are modified in evolution and disease, will require us to know more about what “good enough” is in biology. One conclusion that is emerging from these new studies is that “good enough” may be rather easy to achieve. These features, rather than reducing our respect for biological systems, should makes us even more impressed by what biological systems have achieved. They have achieved facile and flexible ways of achieving their goals quickly with perhaps many routes to success (and fewer routes to failure). This next era of quantitative transcription studies will penetrate far deeper into biological understanding and, though prediction will be more difficult than might have been thought or presently advertised, the answers will in the end be deeper and ultimately more revolutionary.
Hi, I liked your comments – indeed, determination of the grammatical rules that underlie transcriptional enhancers has continued to be a challenging and important problem. Quantitative measurements, such as those conducted by the Berkeley group, provide the basis for thermodynamic models that can reveal general principles of enhancers:
Fakhouri WD, Ay A, Sayal R, Dresch J, Dayringer E, Chiu C, Arnosti DN. (2010) Deciphering a transcriptional regulatory code: modeling short-range repression in the Drosophila embryo. Molecular Systems Biology. 6:341.
Thank you for the comments and for the reference!