Our new Science paper

September 17, 2010 • 7:31 am

Publishing in the journals I call “magazines”—Science and Nature—has become the goal of every graduate student.  A paper in those places is seen as the key to professional success, and I suppose it does help.  But for some reason publishing in magazines doesn’t excite me nearly as much as it does the kids.  I was brought up in an era when there was not so much competition to publish, and when one’s reputation was made by producing a whole corpus of good work.  Where that work appeared was not so important.

I suppose that my Ph.D. advisor, Dick Lewontin, published three or four hundred papers in his career, but I can’t recall one in Science or Nature.  Ditto for his advisor Theodosius Dobzhansky, although I seem to remember one Science paper of his.  In those days you’d succeed if you published good work in Genetics or Evolution.

Do I sound like an old fogey? I guess I am.  All this is a preliminary disclaimer to introducing our Science paper that came out yesterday (see also the news report in Nature).  I can blame the venue on the first author, my terrific graduate student Daniel Matute.  Daniel likes to have fun, but he’s never far from work:

What is the paper about?  It deals with the genetics of hybrid inviability in Drosophila flies.  I’ll describe what we found, and will be as brief as I can given the technical nature of the work.

The problem of how evolution can produce sterile (or inviable) hybrids is an old one.  It occupied Darwin, for instance, in The Origin.  The question that puzzled early evolutionists is how natural selection can “select for” the sterility of hybrids between species. Sterility and inviability, of course, are maladaptive traits, so how can they be produced by natural selection?

With the wisdom of hindsight, we can see that the answer is obvious: sterility/inviability is not selected for directly, but is simply a byproduct of evolutionary changes—adaptive or otherwise—in isolated lineages.  When two evolutionarily isolated lineages have undergone separate and sufficient evolutionary divergence, their genes no longer work well when they’re put together in a hybrid.  And this malfunction can make those hybrids sterile or dead.  That seems simple enough, but it eluded workers for many years after Darwin.

During the Modern Synthesis, both Theodosius Dobzhansky and H. J. Muller made explicit genetic models of how natural selection, for instance, could produce sterile or inviable hybrids.  Although the model was also suggested much earlier, in 1909, by William Bateson, it’s become known as the “Dobzhansky-Muller”, or “DM”, model of hybrid dysfunction.  For those interested in the historical background, see the Genetics paper by my first student, Allen Orr.

The underlying genetic model is simple.  Suppose you start with a single ancestral species, assuming evolution at two genes, gene “A” and gene “B”.  Each gene is fixed for one allele, so the ancestral genotype is aabb.

In one lineage evolution occurs at the “A gene,” with allele A replacing allele a:

Lineage 1:  aabb —> AAbb

In the other lineage evolution occurs at the “B” gene, with allele B replacing allele b.

Lineage 2: aabb —>aaBB

Then all you need to suppose is that in the hybrid between these two lineages, the A and B alleles interact to cause problems, either sterility or inviability:

Hybrid:  AaBb.  Combination of A and B alleles causes dysfunction.

Note that although I use capital letters for the new alleles, there is no assumption about dominance or recessivity.  Note also that the evolutionary change need not have been caused by selection; it could just as easily have been produced by genetic drift or other “nonadaptive” forms of genetic change.

The key insight here is that sterility or inviability of hybrids results from a genetic interaction between alleles at different loci (we call such interactions “epistasis”).  The problems in hybrids stem from combinations of genes that have never “seen” each other in evolution until they co-occur in a hybrid.

That’s pretty simple, but it took an explicit model to ram it home.  Remember, though, that this was still a hypothesis.  There wasn’t much direct evidence for it, and there are other processes, not involving epistasis, that can cause hybrid dysfunction (chromosomal rearrangements are one).

Later work by Allen Orr and his colleagues showed that if the “epistasis” model of hybrid dysfunction is correct, it makes an explicit prediction: the number of genes involved in these deleterious interactions should “snowball” as time passes, increasing exponentially rather than linearly with divergence time.  For dysfunction caused by two-gene interactions, the number of genes involved should go up as the square of the time (t) since the species diverged: t2.  That is, if species have been diverged for two million years, there should be four times more genes causing hybrid problems than when they had diverged for only one million years. (If the bad interactions involve three genes, the number goes up as the cube of time, and so on.)  This more-than-linear increase simply reflects the fact that the process involves interactions between new alleles, and as those alleles accumulate the number of interactions between them goes up exponentially.

So you can test the DM model if you have a way of estimating how many genes cause problems in hybrids, and know the relative divergence times of the species you’re testing.  This is the subject of the Matute et al. paper.

What we did was use two pairs of Drosophila species of known divergence times.  One pair diverged roughly 5.4 million years ago, the other about 12.8 million years ago (we estimated these ages using DNA divergence and the assumption of a molecular clock).  The divergence times between these species then differed by a factor of 2.4 (12.8/5.4).  If the DM model was correct, then the number of genes involved in hybrid problems (we measured the inviability of hybrids) should differ by a factor of at least 5.8 (2.42).

We then estimated the number of genes causing inviability of hybrids in both pairs of crosses using a method I developed years ago, a method based on “deficiency mapping” pioneered by early fly geneticists.  This was laborious, and I’ll omit the messy details, which are given in the paper.

The upshot was that in the younger pair of species, about 10 genes were responsible for hybrid inviability.  In the older pair, about 65 genes were responsible.  Since the pairs differed in age by a factor of 2.4, the number of genes involved went up much faster than the divergence time. In fact, the relative number of genes involved, 6.5, was pretty close to the estimate of at least 5.8.  In other words, the number of genes causing hybrid problems “snowballed” over time, as predicted if those hybrid problems were due to deleterious epistatic interactions between diverged genes.

A paper by by Leonie Moyle and Taluya Nakazato in the same issue (these were published back to back because, fortuitously, another group was working on the same test, but in the plant genus Solanum) showed similar results.  The number of genes  involved in seed sterility “snowballed” over time, but this effect wasn’t seen for pollen sterility.

Together, the two papers provided a nice confirmation of a prediction from the first explicitly genetic model of speciation.  The results do suggest that hybrid problems are the accidental byproducts of evolutionary divergence in separate lineages.

As always, more work is needed.  In flies, for example, we tested only two species pairs, and it would be nice to have a whole array of different species pairs, all differing in divergence times, so you could plot how the genes involved in dysfunction increase over a wide span of time.  That will be very difficult, at least in flies, because the method we used to estimate gene number requires genetic tools available in only one species. But other methods are possible: Moyle and Nakazato used a different one.

The nice thing about this result, from the standpoint of an old fogey, is that my newest student tested and confirmed a population-genetic model of speciation devised by my first student.  The circle is unbroken!


Matute, D. R. I. A. Butler, D. A. Turissini, and J. A. Coyne.  2010.  A test of the snowball theory for the rate of evolution of hybrid incompatibilities.  Science 329:1518-1521.

Moyle, L. C. and T. Nakazato.  2010.  Hybrid incompatibility “snowballs” between Solanum species.  Science 329:1521-1523.

17 thoughts on “Our new Science paper

  1. So, Dobzhansky to Lewontin to you to your student. 4 degrees of separation.

    Kevin Bacon could probably do better, though. 🙂

    Thanks for the nice explanation of a technical paper. Whenever I hear people complain that scientists speak in language others don’t understand, I always reply that they have to in order to get anything done.

    1. It’s better than that—Dobzhansky worked in the lab of Thomas Hunt Morgan, the founder of modern transmission genetics.

    2. BTW, however….

      1. Lewontin RC, Caspari EW. Developmental Selection of Mutations. Science. 1960 Dec 2;132(3440):1688-92.

      You were probably in 5th grade at the time, which is why you didn’t notice.

  2. Ah–real science. A refreshing change after grubbing around in endless religious discussions.

    Excellent to see a prediction so solidly confirmed by experiment! Now that I’ve read the easy version, I’ll go see if I can follow along with the real thing (gambarimasu…).

  3. Great stuff Jerry. I can now cite this in my first year lecture on speciation in three weeks time! And congrats on getting into one of the magazines (not for the first time, I note!)

  4. Thanks for the tip, biology is endlessly fascinating!

    after grubbing around in endless religious discussions

    We can grub around some more: this falsifies Miller’s quantum woo.

    In Miller’s eyes human are special, so they are the only (AFAIU) subjects to his religious magic. Miller predict linear divergence.

    Coyne et al predict and test successfully for exponential divergence. AED (Atheism Erat Demonstrandum).

    1. “In Miller’s eyes human [sic] are special, so they are the only (AFAIU) subjects to his religious magic. Miller predict linear divergence.”

      Predict linear divergence for humans that is, the species that is exclusively under his gods control.

      I also must assume that the divergence alleles are completely orthogonal between the evolutionary population and his fully controlled one. A reasonable assumption I believe, since the point is that the former is under environmental control and the later under artificial control, some of which must resist the environment (or else be unnecessary).

  5. The Modern Synthesis, Theodosius Dobzhansky and H. J. Muller?? … Publishing in Science?

    Gee… I thought you were supposed to have ” a simplistic view of science” and ” a parochial insider’s view of science that does not seem adequately informed by its history, philosophy, and an awareness of how science works…”. 😀

    PS – Well done!

  6. It’s always a shame when interesting papers are stuck behind paywalls, but I suppose laypeople like myself don’t really have the science skillz to analyse it. In any case, sounds like an interesting result and I shall read the H Allen Orr paper to try and grasp the significance.

  7. Congrats on the publication!

    Speaking as a molecular biologist, we often rank Science & Nature publications on par with those of Cell. But alas, in the end, come Science of other, our venue is contingent on public/editor zeitgeist.

  8. Awesome work. Congratulations

    Not to nit-pick, but I think your student would appreciate it if you corrected the typo in the article citation at the bottom of your post — not “Mature” but “Matute.” I suspect you can blame either spellcheck or autocorrect.

Leave a Reply