I’ve been getting swamped with links to this hot article, “Evolution reversed in mice,” including one from my brother (hi, Mike!). It really is excellent and provocative and interesting work from Tvrdik and Capecchi, but the news slant is simply weird—they didn’t take “a mouse back in time,” nor did they “reverse evolution.” They restored the regulatory state of one of the Hox genes to a condition like that found half a billion years ago, and got a viable mouse; it gives us information about the specializations that occurred in these genes after their duplication early in chordate history. I am rather amused at the photos the news stories are all running of a mutant mouse, as if it has become a primeval creature. It’s two similar genes out of a few tens of thousands, operating in a modern mammal! The ancestral state the authors are studying would have been present in a fish in the Cambrian.
I can see where what they’ve actually accomplished is difficult to explain to a readership that doesn’t even know what the Hox genes are. I’ve written an overview of Hox genes previously, so if you want to bone up real quick, go ahead; otherwise, though, I’ll summarize the basics and tell you what the experiment really did.
First, to get through this, here’s the minimum that you need to know. Hox genes are found in clusters, with the genes expressed colinearly—that means that the Hox gene at one end of the cluster is expressed at one end of the animal, the Hox gene at the other end is expressed at the other end of the animal, and the genes in between are expressed in a similarly orderly pattern that maps in sequence along the body of the animal. This paper only looks at Hox1, the anteriormost Hox gene, which is expressed far forward in the neural plate of the mouse, in cells that will contribute to anterior parts of the brain. You can forget about Hox2, Hox3, … Hox13 for now.
However, we do have to deal with multiple copies of Hox1. About half a billion years ago, there was a massive set of gene duplications in the vertebrate lineage—we have four clusters of Hox genes. Each of these four copies is designated by a letter, a-d. Hoxa1 is the first Hox gene in the first bank, Hoxa2 is the second Hox gene in the first bank, etc., and Hoxb1 is the first Hox gene in the second bank, and so on. This diagram will help you sort them out.
Again, though, for this paper we only care about Hox1, so we are only going to consider the
four three copies at the left side of the diagram, Hoxa1, Hoxb1, Hoxc1, and Hoxd1. You may notice that there is no Hox1 gene in the c cluster; it was lost long ago, and no mammal has a copy any more. Furthermore, the Hoxd1 gene seems to have a negligible role in patterning the brain, so we can ignore it. This experiment is looking only at Hoxa1 and Hoxb1.
So: two paralogous genes, Hoxa1 and Hoxb1, expressed in similar regions at the anterior end of the nervous system, and conserved over half a billion years of evolution. They must do different jobs in development, or evolution would tolerate deletions of one or the other of the genes. What exactly are the specialized roles of these two genes?
We know that Hoxa1 is essential; delete it, and the hindbrain forms abnormally, with certain regions enlarged at the expense of others, disrupting the breathing centers of the brainstem, and you get a mouse that dies right after birth. The two genes also differ in protein sequence (they have 49% amino acid identity), and they also differ in significant ways in the timing and pattern of their expression in the brain. The difference in sequence suggests that Hoxa1 and Hoxb1 could have different functional roles, and the difference in how they are expressed suggests that they could be different in their regulation. Hey, this looks like a system where we can explore a big question in evo-devo: are changes in the coding sequence of proteins the most significant factor in evolution, or are changes in the non-coding regulatory regions more important? (Of course, we expect that both will be important, and there will be differing degrees of importance in different genes…but this is a case where we can directly compare these two paralogous genes and assess which differences are most important here.)
The experiment is simple to explain but tricky to do (the Capecchi lab, though, has developed some of the most amazing tools in molecular genetics): it’s a gene swap. Pluck out the coding sequence for the Hoxa1 gene, and put it in place of the Hoxb1 gene, where it is surrounded by all the Hoxb1 regulatory elements. Also pluck out the Hoxb1 gene, and imbed it in place of Hoxa1. This means that Hoxa1 will be turned on at the times and places appropriate for Hoxb1, and vice versa. If the function of the coding sequences of these two genes are distinct and incapable of substituting for one another, we’d expect the resultant mouse to be thoroughly messed up.
The answer? The mutant mice are normal.
This means that as far as we can tell, the Hoxa1 and Hoxb1 proteins are functionally interchangeable. The differences that they’ve accumulated in the peptide sequence over that half billion years of evolution do not significantly change how either protein works, and all the important action must have taken place in the evolution of the regulatory control of each.
Let’s take a look at the evolution of the regulation of these two genes. In the diagram below, the figure to the left illustrates the ancestral state, just after that ancient gene duplication. That distant gnathostome ancestor has four identical copies of the Hox1 genes, A1, B1, C1, and D1. They all also have the same regulatory elements, the chief of which are called ARE (the diamond) and RARE (the circle). RARE is the Retinoic Acid Regulatory Element—these genes are triggered to be expressed by retinoic acid. ARE is the Auto-Regulatory Element. Once the protein is expressed, it feeds back and binds to its own DNA, promoting the production of more copies of the gene.
The right side of the figure shows the changes that occurred in vertebrate evolution. The A1 gene has a regulatory specialization: it is more sensitive to retinoic acid (the larger RARE component), but is less sensitive to autoregulation (the smaller ARE diamond). The B1 gene has gone the other way, and while less sensitive to retinoic acid, is more strongly self-activated. C1 is gone altogether, and D1 is only weakly regulated. What you need to make a normal mouse is one gene that is readily activated by retinoic acid, which then binds to the ARE site of both Hoxa1 and Hoxb1, and another gene that is sustained by that autoregulatory activity. Lose either one, and you don’t get the full dose and range of timings necessary for normal development; however, the system isn’t too picky about which Hox1 gene is regulated by these two patterns.
The way they know this is by careful analysis of multiple kinds of gene swaps, and by studying mutant homozygotes and heterozygotes combined with deletions. Not all of the mice resulting from these experiments are happy; some are hypermorphs, where the Hox1 genes are too active, leading to reduced viability, while some are hypomorphs, where the genes aren’t active enough, leading to deficient differentiation of nervous tissue and facial paralysis. They also identified a functional difference between Hoxa1 and Hoxb1 (they aren’t completely interchangeable!): Hoxb1 is better at activating the autoregulatory region than is Hoxa1.
This wonderful diagram encapsulates all of these various experiments.
Start at the top, with the wild type animal. The normal sequence of events (simplified here) is that the retinoic acid signal turns on A1, which then switches on B1 via its autoregulatory element. B1 then strongly turns on its own ARE, producing a normal level of expression.
The second diagram is an experiment where the B1 peptide has been replaced with the A1 peptide. The initial sequence is the same, A1 is turned on by retinoic acid, and it switches on the ARE…but that produces an A1 protein, which doesn’t autoregulate as well, so you get a weaker overall level of expression, and produce a partially paralyzed hypomorph.
The third experiment is the complement: this animal only makes B1 peptide. Retinoic acid turns it on, and it strongly activates the next protein, which also strongly turns itself on, and the combination of two strongly autoregulating agents leads to an overexpression of the Hox1 gene.
The fourth experiment is the double swap, with B1 in A1’s place and A1 in B1’s place. This produces an essentially normal animal, because it has the appropriate balance of one weak activator of autoregulation and one strong activator of autoregulation.
The last part of the diagram finally gets to the experiment that the news is making so much of, the reconstruction of the primitive state. The way the two Hox1 genes have specialized is that one has developed a robust RARE to make it sensitive to retinoic acid, and the other has a strong ARE to promote solid autoregulation. The ancestor before the gnathostome duplication event had only one Hox1 gene, but it would have had both regulatory components. So in this last experiment, they destroyed the B1 gene, and spliced the B1 ARE to A1, producing a hybrid single gene with both regulatory components. This situation also produces a normal mouse!
What the paper actually shows is not how to reverse evolution, but that the primitive chordate condition was to have a single, multifunctional gene, and that the result of a duplication event was a specialization of each copy to support separate subfunctions. What it also is is an example of is an increase in complexity, and the evolution of an irreducible system—one where each of two gene products is essential, and the animal dies without either—from one with a single generalized gene. While the research doesn’t reverse evolution for an organism, it does turn back the clock on a single gene and show us key changes in its regulation.
Tvrdik P, Capecchi M. (2006) Reversal of hox1 gene subfunctionalization in the mouse. Dev Cell 11(2):239-50.