Intelligent Design creationists are extremely fond of diagrams like those on the right. Textbook illustrators like them because they simplify and make the general organization of the components clear—reducing proteins to smooth ovoids removes distractions from the main points—but creationists like them for the wrong reasons. "Look at that—it's engineered! It's as if God uses a CAD program to design complex biological systems!" They like the implication that everything is done with laser-guided precision, and most importantly, that every piece was designed with intent, to fill a specific role in an apparatus that looks like it came out of a high-tech machine shop at a Boeing aerospace lab.

This is, of course, misleading. Real organelles in biology don't look glossy and slick and mechanical; they look, well, organic, with fuzziness and variability and, most importantly, mistakes and slop. What these biological machines look like is not the precisely engineered output of a modern machine shop, but like bricolage. Bricolage is a term François Jacob used to contrast real biology with the false impression of nature as an engineer. It's an art term, referring to constructions made with whatever is at hand, a pastiche of whatever is just good enough or close enough to the desired result to make do. It covers everything from the sculptures of Alexander Calder to those ticky-tacky souvenirs made from odd bits of driftwood and shells glued together that you can find at seashore gift shops.

The closer we look at the developmental biology of organisms, the more apparent the impromptu, make-do nature of their construction is. This is not to imply that they don't work well or efficiently, but only that the signature of intent is missing. What we see is function cobbled together out of scrap from the junkyard. One clear example of this property is a gene, nanos, in Drosophila.

First, a brief review. Previously, I wrote about maternal effect genes . Some genes, the maternal effect genes, are expressed in the mother's genome and are essential for packaging information into the egg before it is laid. The example I used was bicoid, a gene that is expressed in the mother fly's ovaries and which produces a gene product that is stored in high concentration at the front end of the egg, and at low concentration at the back end. When the embryo is developing, cells can detect the concentration of bicoid and turn on their own, zygotic genes accordingly. The first zygotic genes in the fly are called the gap genes, and they turn on in discrete bands along the length of the embryo. For instance, one gap gene is called hunchback, and it is turned on only where the bicoid concentration is high, at the front end, and it is not turned on at all in the back half of the embryo, and there are other genes that are only active in the back half, and are turned off in the front.

The story is a little more complicated than that. There is this lovely gradient of the bicoid morphogen that is essential for specifying the front end of the animal; lose it by mutation, and the embryo can't make a head. However, there is also a posterior gradient, a substance that is in high concentration in the back and low in front, that is essential for specifying posterior structures. That substance is called nanos, another maternal effect gene. Flies with a mutant nanos gene produce embryos that lack abdomens.

So, the fly actually has two complementary gradients, one of bicoid and another of nanos, both essential for determining the anterior-posterior organization of the animal. The fly invests a lot of effort into setting up these gradients; there is an army of genes (staufen, valois, oskar, vasa, tudor, to name a few) dedicated to making sure that the nanos gradient is set up properly, and another, smaug, that works to prevent any nanos that leaks into the wrong places from getting expressed. Obviously, nanos is mission-critical stuff that has to be in exactly the right place at the right time. What does it do?

I've already explained what bicoid does: it's a transcription factor that binds to DNA and turns genes off and on. Nanos is a little bit weird. All it does, as far as we know, is bind to hunchback RNA and modify it so that it can't be translated. That's it. It's an inhibitor of hunchback activity.

The reason the fly needs that is also illustrated in the diagram above. There's bicoid high at the anterior end, and nanos high at the posterior end, and...hunchback all over the place. Remember that I told you that hunchback is a zygotic gene that is turned on in the embryo in response to bicoid? That's true, but there's another complication: the mother fly also packs the egg with a low level of maternal hunchback RNA. We don't know why. It seems rather useless, and actually interferes with normal development. If nanos is not present, it is the elevated levels of hunchback at the posterior end of the embryo that confuses cells there, and makes them develop into more anterior structures. What the nanos protein does is illustrated here—it simply purges hunchback from the posterior end of the embryo.

There's a telling experiment that reveals how trivial the function of nanos is. We can make flies in the lab that lack maternal nanos. We can make flies that lack maternal hunchback. We can make flies that lack both maternal nanos and maternal hunchback, and here's the kicker, these flies produce embryos that develop completely normally. Maternal hunchback RNA is a mistake, a sloppy bit of unnecessary secretion that does nothing for the embryo, and maternal nanos is an elaborate, Rube-Goldberg mechanism that has been patched in to correct the stupid mistake.

Now if I were an intelligent designer building a fly from scratch, and I saw this little design defect, the way I would correct it is the obvious one: I'd fix the fly ovaries so that they weren't dribbling a completely useless protein into the egg. I wouldn't assemble a complex of a half-dozen proteins that were integrated into the cytoskeleton and pump a hunchback-suppressor to the right place, with other proteins floating around to make sure my hunchback-suppressor didn't do damage in the places where hunchback is supposed to be turned on.

As design, the nanos posterior gradient simply doesn't make sense. As bricolage, however, it's perfectly reasonable. Evolution does not demand a best or elegant solution, only one that works well enough. We suspect that other insects have more essential gradients of posterior positional information, and that Drosophila has inherited this mechanism from ancestors that relied more on it. Flies have some oddities to their development, though, and they've come to depend more and more on the bicoid gradient, and the posterior gradient system has been declining in importance. Nanos localization is a module that's been sitting in the Drosophila junkyard, and now it persists to compensate for a little slack in the mRNA packaging system of the ovaries.