Pair-rule genes


The general pattern of developing positional information in Drosophila starts out relatively simply and gets increasingly complicated as time goes by. Initially, there is a very broad distribution of a gradient of a maternal morphogen. That morphogen then triggers the expression of narrower (but still fairly broad) bands of aperiodic gap genes. The next step in this process is to turn on sets of genes in narrow, periodic bands that correspond to body segments. This next set of genes are called the pair-rule genes, because they do something surprising and rather neat: they are turned on in precisely alternating bands. In the picture above, for instance, one pair-rule gene, even-skipped, has been stained blue, and it is expressed in parasegment* 1, 3, 5, 7, etc. Another, fushi tarazu, has been stained brown, and this gene is turned on in parasegments 2, 4, 6, 8, etc.


Compared to the gap genes, the pair-rule genes are a nightmare of complexity and special rules. The gap genes have elaborate interactions among themselves to set up their domains of expression, and the gap genes in turn activate and repress the primary pair-rule genes (even-skipped, hairy, and runt), and they in turn interact with each other and with a set of secondary pair-rule genes (fushi tarazu, odd-skipped, paired, sloppy-paired). Some overlap in expression and activate each other, others have complementary regions of expression and inhibit each other. As the grossly simplified diagram to the left shows, even-skipped (eve), for instance, up-regulates hairy (h), but wherever eve is turned on, it turns off runt and fushi tarazu (ftz).

Oh, that biology were actually so simple.

I’ve been reading a paper on the regulatory logic of the pair-rule genes by Sanchez and Thieffry (2003); it’s an interesting combination of thorough review of pair-rule interactions, and an attempt to model those interactions as logical formalisms. The model is derived from observations of the pattern of gene expression and epigenetic experiments that track regulatory interactions by knocking out genes or overexpressing them.


Here, for example is a summary of the segmental pattern of these genes in the early embryo (top half) and late embryo (bottom half). You can see that we go from a very broad, fuzzy pattern for each to a much tighter, harder-edged band. (OK, it’s a cartoon, so you can’t really see it there—so look at the photo to the right of eve and ftz in the late embryo. Sharp! Cells express it or they don’t, and they lie in discrete lines.)

Schematic representation of the main pair-rule gene expression domains. The initial expression patterns of pair-rule genes form seven bell-shaped domains (top), which are later refined into final narrow stripes (bottom). Gene hairy is shown on the top of the figure because this gene ceases its expression earlier than its target genes ftz and run, before the final refinement process leading to the narrow stripes of the pair-rule genes directly involved in en and wg activation. Five different regions (I–V) defining two parasegment borders are indicated. The interactions between the pair-rule genes are the same in all locations along the anterior–posterior axis where the parasegment boundaries, represented by thick lines. These are flanked by two regions expressing wg (region I or IV) and en (region II or V) respectively. Regions IIIa and IIIb correspond to the middle of the parasegment, each representing a region adjacent to the en or wg expressing cells within the parasegment. As a whole, region III is characterized by the lack of expression of both en and wg and by the expression of odd. The primary and secondary pair-rule genes are shown in graded shadow and empty boxes, respectively. Genes symbols: runt (Run), even-skipped (Eve), fushi-tarazu (Ftz), odd-skipped (Odd), paired (Prd), sloppy-paired (Slp), engrailed (En) and wingless (Wg).

Very nifty, huh? We’re not done yet, though. The authors have put together a diagram to show all the logical interactions, negative and positive, in this system:

Final pair-rule expression domains after the refinement process and interactions between these genes. Solid arrows stand for positive interactions, dotted blunt arrows indicate negative interactions.

Now that’s the kind of mess you get with a few million years of evolutionary tweaking. The idea is that there is a “pair-rule module”, a simpler underlying scheme that inhibits complementary genes to generate discrete, exclusive domains of expression. What evolution does, though, is spawn duplicate copies that then are free to acquire a slightly different regulatory logic…and what you end up with is an elaborate assortment of partially redundant, overlapping units, each with minor variations. Drosophila has gone a long way down the road towards hard-coding specialized functions into the regulation of these genes, and the general module is hard to pick out anymore.

One reason developmental biologists have become increasingly interested in looking at these genes in other organisms is the idea that common features between flies, beetles, grasshoppers, fish, mice, etc. will help us more clearly see what are the core modules and what are the late-evolving rococo additions.

Oh, and yes, we are finding these genes in other organisms. I’ve mentioned hairy homologs before as key regulators of segmental organization in vertebrates.

*One somewhat confusing issue I won’t get into now is the segment/parasegment distinction. Morphological segments are the distinctly bounded regions you can see in the insect’s body; think of the bands of cuticle you can see in the adult fly’s abdomen, for instance. Development is always throwing odd discoveries at you, though. It turns out that the fundamental genetic/molecular unit is actually half a segment out of phase with the morphology, hence it is called a parasegment.

Sanchez L, Thieffry D (2003) Segmenting the fly embryo: a logical analysis of the pair-rule cross-regulatory module. J Theor Biol 224(4):517-537.


  1. llewelly says

    One somewhat confusing issue I won’t get into now is the segment/parasegment distinction. Morphological segments are the distinctly bounded regions you can see in the insect’s body; think of the bands of cuticle you can see in the adult fly’s abdomen, for instance. Development is always throwing odd discoveries at you, though. It turns out that the fundamental genetic/molecular unit is actually half a segment out of phase with the morphology, hence it is called a parasegment.

    Who designed this mess again?

  2. Mary says

    See? This is definitive proof that there is a God, and that He hates scientists. It’s just that they keep on “finding” rules about these so-called “genes” that make up these “organisms,” as you call them, instead of accepting that the only important part is that God made it all. Jeez.

  3. says

    Interesting… because I realized about 20 years ago that the amazing patterns of animal colouration, from a moth’s wing to a trout’s speckles, could be created by alternating light-dark bands of pigment intersecting at different angles. Presumably if this works for more complex systems it could also work for something simple like color.

  4. Arnosium Upinarum says

    “Oh, that biology were actually so simple.”

    Wow. “Very nifty” indeed! I’m not a biologist, and this stuff is hard for me to wrap my head around (I’m slow but stubbornly get there) but I am awestruck every time biology actually details some of the complexity behind development like this, not to mention “the kind of mess you get with a few million years of evolutionary tweaking” (indeed “rococo”), that now nearly conceals the underlying regulation module in this Drosophila case.

    There it is. We can LOOK at it. We can SEE it, right there, in all its complicated glory. Its beautiful. (And its beautiful work by Sanchez and Thieffry.).

    I love this site. Learn something amazing every day! Makes one dig for more. Thanks again PZ! (And for the links!).

    But then, out of the blue (undoubtedly triggered by the marvelous micrographs of the stained segment patterns in those embryos), I found myself comparing this meticulous work with the fashionably-dunderheaded and gross political accounting of “Red vs. Blue states” these days. Then I considered that our societies are the product of mere thousands of years of social-evolutionary tweaking, and I found myself wishing: Oh, that sociology were actually so simple as BIOLOGY.

    Downright depressing.

    I’m not a sociologist either, but can social science be that hyper-byzantine complex, or is it just because nobody can look at an ‘image of behavior’ without resorting to a graphic or chart? Its based on “observation” (often anecdotal) with no images in sight – its all statistical and systems, yet the maths applied to it have not come anywhere near coping with it. With the vast hodgepodge of “independent minds” running around “regulating the expression” of societies, I suppose its no wonder the social sciences haven’t yet even produced a single unifying theme or set of foundations, as physics and biology have.

    Society is evolution run amok, with significant changes now happening globally by the hour. And the rate of change is increasing. What are we LOOKING AT – FUNDAMENTALLY – when we look at this rapidly-changing apparition?

    The most depressing part is that its arguably the area which is the most urgent to our collective survival…

  5. LeeLeeOne says

    PZ… there’s no way I could understand anything from the above. Biology is not my forte. I only clicked on this because I saw the first picture and thought, “what the hell is a picture of my front door mud rug doing on pharyngula?!”

    Ok, now I’ve proven myself to be a total infant when it comes to attempting to understand complex systems.

    I’ll finish my glass of wine, and attempt to remember where in the attic I placed my freshman text books…

  6. Happ says

    PZ Meyers you are an idiot! You have to conjujate the quantam anomalies when dealing with such unpredicatable variables. The diffusion flux is just completely unstable. Where are your parameters! The idea of the complex systems of the physics of the principles hardly make any sense. What are you saying?

  7. Dan K. says

    Waiter, there’s a fly in my soup!

    Of course, sir, our soup is good to the last drosophila.

  8. Loren Petrich says

    And as PZ had explained in another blog post, this hardcoding of segmentation is largely a feature of certain of the insects, the “long-germ” ones, whose segumentation develops all at once. The less-derived approach is the “short-germ” one, where segments emerge from the rear end of the growing embryo. Not only most other arthropods, but also annelids and vertebrates also use that approach, so if segmentation is ancestral to arthropods, annelids, and vertebrates, then that ancestor would likely have done that.

    However, the long-germ approach has the advantage of allowing quicker growth, which may indicate why it has been selected for in some cases, despite its greater complexity.

  9. says

    Wow. I have to say that developmental biology was not one of my favourite undergrad courses, largely due to a very dry lecturer. This was also before I discovered how interesting transcriptional regulation can be. So it’s great to read about the pair rule genes again with a bit more understanding and appreciation of the nature of gene expression regulation. Very cool stuff, thanks!

  10. RavenT says

    Development is always throwing odd discoveries at you, though. It turns out that the fundamental genetic/molecular unit is actually half a segment out of phase with the morphology, hence it is called a parasegment.

    And morphological units can be out of sync with each other in the same way–vertebrae are not formed out of each somite in a neat one-to-one and onto pattern, but rather from the last half of one somite, plus the first half of another somite (hi, Billie!).

    Teasing out the finer details of this process is just so fascinating! Both in its own right, and in its potential applications–we are right on the brink of really gaining some useful knowledge about what underlies certain diseases, and maybe being able to do something about them.

    I genuinely don’t understand why some people try to sabotage this knowledge…

  11. vincent fleury says


    I am a scientist interested in biomechanical aspects of pattern formation; although I am not that aware of the details of all the chemical feedback loops of the stripes generation in drosophila, I am very much skeptical about the overall point of view used to describe this situation.
    Indeed, first of all, as a scientist of pattern formation, you are (I am) interested in the first place by patterns,

    but what do we see in the drosophila at the stage described here? : something which already has a shape (rugby like). This shape is not discussed. Next, we see stripes, which obviously do not impart any modification on the shape, while the shape itself already imparts a shape to the stripes (boundary condition effect) : they bend towards the edge, not only they bend towards the edge, but from the picture above, we see that the wavelength is smaller in the concave side, than in the convex side, which is another evidence for boundary effects, and possibly stress related effects

    In addition, we look at the stripes, and we obviously see rows of aligned cells, with some noise, and the noise is reduced as time passes by. The stripes are not, from what I see, chemical stripes, but more stripes of cells, which contain certain chemicals. That makes a huge difference, because cells do not diffuse. They move by micro-displacements, i.e. the final distribution of cells is the result of the integral over time of the deformation rate, which is a tensorial field, and this has nothing, or only indirectly something to do with reaction diffusion of chemicals. In addition, rows become more regular with time. This is exactly what is observed by pushing boundaries against each other, which eventually become more regular (as peas in a box, cigarettes in packets, or bllod vessels in the yolk sac, see Thi Hanh Nguyen et al, Dynamics of vascular branching morphogenesis The effect of blood and tissue flow Phys rev E 2006, 73 61907

    Nothing of all these microdisplacements which add up to position cells is mentionned in the modelling (but may be thay are mentionned in the orginal articles, I do not know).

    Not even mentionning that the diagramms of gene induction regulation lacks any possible stress related effect, which is liley from the arguments given above. In the end, I think that one should still be very cautious with the chemical models, especially in relation to pattern formation. They may aswell be completely wrong.

    Best wishes

    Vincent Fleury