Over the last few years, scientists have figured out how to recreate biological molecules that were last seen on Earth hundreds of millions of years ago. Until now, scientists have reconstructed ancient proteins to gather clues about life was like long ago. But now some scientists at the University of Oregon have done something new with these old proteins: they used them to figure out how evolution produces complex systems–exactly the sort of systems that creationists would have us believe cannot evolve.

Scientists reconstruct an ancestral protein by tracing its evolution into new versions carried by living species. Along each lineage, the gene for that protein picks up mutations, some of which alter the structure of the protein. Scientists can determine many of those mutations, and by working backwards up the evolutionary tree, they can determine what the original gene looked like. Thanks to powerful statistical techniques, they can determine how much confidence they can have in each letter in the genetic sequence they reconstruct. If they find a lot of statistical confidence in the overall sequence, they can then go to the lab and use it as a guide to build the corresponding protein. And once they have the protein in hand (or in beaker), they can see how it works. In this 2004 paper, University of Oregon biologist Joe Thornton reviewed the latest advances in this molecular resurrection. Scientists have recreated one of the light-gathering proteins from the eyes of the common ancestor of birds, crocodiles, and dinosaurs that lived 240 million years ago. The protein still catches light in its resurrected form–and turns out to be particularly good in dim light, suggesting that this ancient reptile was a nocturnal creature.

Thornton and his colleagues have now used this same method to learn about the evolution of signals our cells use to talk to one another. In particular, they look at hormones and the receptors on the surface of cells to which they attach. Hormones coursing through our body maintain harmony between our cells, keeping levels of various molecules in balance while allowing our bodies to respond quickly to challenges–the most famous example being the fight-or-flight rush of adrenalin.

This communication can be ruined by misunderstanding. “Did you say boost glucose levels up a little? Sorry–I thought you meant it was time to utterly freak out.” To avoid that confusion, the structure of hormones only allow them to latch onto certain receptors, a bit like keys fitting into certain locks. (Only a bit, though–something I’ll get back to in a little while.)

The genes for hormone receptors in humans and other vertebrates show clear signs of having evolved from common ancestor. Take the mineralocorticoid receptor (MR for short). In our bodies it responds to a hormone called aldosterone that regulates our electrolytes in the kidneys. We’ve got MR, dogs have it, birds have it, snakes have it, frogs have it. In fact, all known land vertebrates have various versions of MR. Even ray-finned fish have it. Even skates (a kind of cartilaginous fish) have it. The MR gene we carry is more like the MR gene carried by frogs than it is to ray-finned fish or to skates. And if you look at two MR genes carried by fish, they’re more similar to each other than they are to ours. But ray-finned fish MR genes look more like ours than they do to those of skates.

This fits perfectly with what the fossil record tells us. The ancestors of skates and other cartilaginous fish branched off from other living vertebrates about 450 million years ago. Ray-finned fish and land vertebrates descend from a more recent ancestor. And all land vertebrates descend from an even more common ancestor. The MR genes reflect that ancestry.

A few years ago Thornton discovered where the original MR receptor came from. He found that MR genes all bear a striking resemblance to another sort of receptor, the glucocorticoid receptor (GR for short). GR responds mainly to cortisol, and is important for coping with stress and infection. It does not respond to aldersterone, the signal for MR.

Thornton hypothesized that MR and GR are the product of an ancient gene duplication. An ancestral gene for a hormone receptor was accidentally copied, and two version of the receptor began to be produced on ancient fish. Over time, each gene acquired mutations that altered how its receptor responded to hormones, and eventually the two versions evolved into MR and GR.

So this left Thornton with a puzzle: how is that that an ancestral receptor 450 million years ago gave rise to two such different sorts of receptors, one (MR) sensitive to aldosterone and used to regulate electrolytes, and the other (GR) an immunity-linked receptor sensitive to cortisol and not sensitive at all to aldosterone?

The puzzle only gets deeper when you consider the evolution of aldosterone, the hormone that attaches to MR. Only land vertebrates like ourselves make it. So it must have evolved after our ancestors branched off from the ancestors of living fish, about 400 million years ago. In other words, the lock evolved 50 million years before the key.

Some might argue that this means that these receptors and their hormones could not have evolved by mutation and natural selection. They only have their own personal disbelief to go on, however. Thornton took another tack: he did some tough science.

Thornton and his colleagues reconstructed the ancestral receptor that gave rise to both GR and MR. They figured out the sequence of its gene by comparing the genes for MR and GR in land vertebrates and fish, and also looking at other related receptors. They focused their attention on a 247-nucleotide-long section of the receptor gene. This stretch encodes the part of the receptor to which the hormone attaches. The scientists ended up with a sequence of DNA that had a 94% probability of being the ancestral sequence. Two-thirds of the sequence had a 99% probability. They focused their attention on a section of the receptor where the hormone attaches, measuring 247 amino acids long. They calculated that their reconstruction had a 94% probability of being correct, with two-thirds of the amino acids getting at 99% rating.

The scientists synthesized a gene with this sequence and inserted it into cells. The cells used the new gene to make the ancestral receptor, which the scientists could then test. They discovered–weirdly enough–that the ancestral protein was sensitive three different hormones: aldosterone (the hormone specific to MR), cortisol (the GR hormone), and a third, called DOC.

It may seem surprising that the ancestral receptor would respond to aldosterone, a hormone that did not evolve until tens of millions of years later. But it’s not so surprising when you compare them to living fish. Living fish don’t make aldosterone, and yet it can still attach to fish MR anyway. Obviously, the fish aren’t making these receptors to snag hormones they don’t make. Instead, it seems that in fish, MR are responding to DOC, which is very similar to aldosterone. In the ancestors of tetrapods, DOC evolved into aldosterone and took on its function that it has in our own bodies.

Here’s where the lock and key metaphor can cause mischief. Receptors are not built out of metal. They’re loops and spirals of atoms that can flex. So a receptor that is adapted to respond to one hormone may have the capacity to respond to another one. Scientists call such molecules promiscuous. The promiscuity of proteins is a big area of research these days. It explains, for example, why bacteria can feed on pollutants that were nearly nonexistent a century ago. Some of their proteins evolved for other functions, but had the potential to be used to eat a new kind of food. Thornton’s work suggests that MR was preadapted to respond to aldosterone, much like fish with fingers were preadapted for walking on land.

But if these three hormones could all latch onto the ancestral protein, why is it that today its descendant receptors are specialized only for certain hormones? MR can still respond to cortisol, but it doesn’t appear to get the chance to. That’s because kidneys and other tissues that produce MR also make enzymes that destroy cortisol.

GR is different. It does not respond at all to aldosterone. So somehow it had to lose its ability to latch to aldosterone after it evolved from the ancestral receptor. Thornton and his colleagues found that GR differs from the ancestral receptor by two significant mutations. The scientists tinkered with the ancestral receptor to see how each of these mutations affected it.

One mutation (called L111Q) was devastating. It rendered the receptor unable to latch onto any of the three hormones (aldosterone, cortisol, or DOC). But the other mutation (called S106P) only reduced the ability of aldosterone and cortisol to latch onto the receptor. Its response to DOC was unaffected. And here’s where things get extremely cool. Thornton took some of these S106P mutant receptors and then added the L111Q mutation. Now the mutation was not devastating at all. The receptors completely lost their ability to respond to aldosterone but recovered their ability to respond to cortisol.

Scientists have discovered this odd feature of mutations in other genes in other species. The effect of a mutation depends on the other mutations that have already affected a gene. Depending on which mutations are already there, a mutation may be harmful, neutral, or even beneficial. So it would have been unlikely that the ancestral receptor could have experienced the L111Q mutation first. Fish that carried it might have lost their ability to respond to any of the hormones. But they could have survived the S106P mutation, and afterwards the L111Q mutation would have had a different effect altogether. Instead of making the receptor useless, the mutation would have fine-tuned it to respond to cortisol but not to aldosterone.

This study took me quite by surprise. I had been so fascinated by resurrected proteins as a sort of molecular Jurassic Park that it never occurred to me that somebody might use them to work out a step by step hypothesis for the molecular changes that can produce complex systems. I can only wonder how many other proteins–perhaps even precursors to the proteins that give us language or autobiographical memory–will return to tell us their stories.

(This paper will appear tomorrow appears today in Science, but here’s a press release from Oregon.)

Update, Friday 4/7: Here’s my take on the Intelligent Design response to Thornton’s work.

Update, 4/10: And the final take. 

Originally published April 6, 2006. Copyright 2006 Carl Zimmer.