What role does the appearance of new genes, versus simple changes in old ones, play in evolution? There are two reasons why this question has recently become important.

The first involves a scientific controversy.  Some researchers—the most prominent being evo-devotee  Sean Carroll—maintain that most important evolutionary change, at least in body form, involves changes in regulatory sequences rather than simple changes in genes themselves, or the appearance of new genes.  This question hasn’t yet been answered, since we don’t know a great deal about those mutations that have been important in creating new body plans.

The second controversy is religious.  Some advocates of intelligent design (ID)—most notably Michael Behe in a recent paper—have implied not only that evolved new genes or new genetic “elements” (e.g., regulatory sequences) aren’t important in evolution, but that they play almost no role at all, especially compared to mutations that simply inactivate genes or make small changes, like single nucleotide substitutions, in existing genes.  This is based on the religiously-motivated “theory” of ID, which maintains that new genetic information cannot arise by natural selection, but must installed in our genome by a magic poof from Jebus.

I’ve criticized Behe’s conclusions, which are based on laboratory studies of bacteria and viruses that virtually eliminated the possibility of seeing new genes arise, but I don’t want to reiterate my arguments here.  What I want to do is point out a new paper by some Chicago colleagues that suggests that new genes, at least in the genus Drosophila (fruit fly), not only arise pretty quickly, but also diverge very quickly to become essential parts of the genome.

The paper, by Sidi Chen, Yong Zhang, and my friend Manyuan Long, appears in this week’s Science: “New genes in Drosophila quickly become essential.”  It’s a clever piece of work.  What the authors did was compare whole-genome sequences between various species of Drosophila (there are now many of these) to see how often new genes appeared in one lineage: the lineage that diverged from the ancestors of D. willistoni to become D. melanogaster.  The divergence between these two lineages is 35 million years, but by comparing the genomes of other species that branched off these two branches, they could estimate how often new genes arise over the entire period from 3 million to 35 million years ago.

What do they mean by “new genes”?  These are genes in D. melanogaster that aren’t found in D. willistoni, but have arisen since their divergence by several processes—most often the duplication of an ancestral gene or its RNA followed by extensive genetic divergence, so that the gene acquires a brand new function.  (This process accounts for about 90% of the new genes.  Some genes, however, are so different between the species that how they arose is a mystery.)  These “new genes,” then, would qualify as what Behe calls “gain-of-FCT” adaptive mutations (“FCT” = functional coded element): the kind of mutations that Behe did not see arising in short-term lab experiments on bacteria and viruses.

Chen et al. found that a surprisingly large number of genes had arisen in the D. melanogaster lineage over this 35-myr period.  Here’s a summary of their results:

• The authors identified 566 new genes that arose over this period. That’s about 4% of the total genes in the D. melanogaster genome.  And that’s quite a few given that the divergence is only 35 myr.  The genus Drosophila itself (including the scaptomyzids) diverged from its sister group about 63 million years ago, so we can estimate that, in the genus as a whole, at least 7% of the genome comprises brand new genes.

• The authors were able to take a sample of these genes (195 of them) and knockdown their transcripts using novel RNAi technology (this involves inserting transposable genetic elements in those genes and then using those elements to kill the genes).  They found that about 30% of these new genes are essential for viability—that is, the fly dies if it has no active copies.  This proportion didn’t vary depending on how long ago the “new” gene had arisen.   Nor did it differ much from the proportion of “old” genes (those present in both lineages) that are essential for viability, which is about 35%.  It seems, then, that even if these genes arise as duplicates from pre-existing genes, they quickly assume new functions that make the fly unable to survive without them.

• The “new function” conclusion is supported by two other pieces of data.  First, the average difference in DNA sequence between the “new” genes in the D. melanogaster lineage and their parental copies (that is, the genes from which they originated, usually by duplication) is 47.3%.  That’s a big difference—a change in nearly every other nucleotide.   Second, there are new ways to determine what the new genes do: by estimating which proteins in the genome each new gene’s protein product interacts with.  Chen et al. found that many of the products of new genes interact with proteins completely different from the ancestral genes.  This implies that the new genes have evolved completely different functions.  And, as theory suggests, that’s the way these genes become essential: at first they do the same thing as their ancestral genes (they’re duplicates, after all), but as they diverge they assume new functions (usually impelled by natural selection) that fit them into new developmental pathways.  In this way a gene that is at first “gratuitious” can become essential. It’s nice that we can actually see this happening with protein-protein interaction data.

• In further support of the above scenario for the evolution of new genetic information, the authors found that in young and new “essential” genes, there was a strong signature of natural selection having acted, as suggested by the high rates of DNA substitution.  As the “new” essential genes become older, and assume new functions, these rates slow down.  This again supports the theory of how new genes originate: when they’re formed by duplication, they are quickly eliminated from the genome (see below) unless they diverge quickly to do something new.  Thus the duplicates that do survive are usually those that have diverged quickly.  Once the new function has been assumed, and the gene is essential, selection then acts to preserve its new function by eliminating new mutations (“purifying selection”).

• These results, which show that new genetic information (“FCT”s) arises quickly, don’t imply that every new gene duplication becomes a brand-new gene with a new function. That’s far from the case.  We don’t know the figure in Drosophila, but in the human lineage it’s estimated that only about 5% of new duplications diverge to become new genes that do something novel.  The rest are inactivated, becoming dead “pseudogenes” that don’t do anything. In Drosophila these are quickly removed from the genome, but in our own lineage many of them linger, so we can estimate the proportion of duplicated genes that don’t go on to do something new.

• Nevertheless, genes duplicate frequently enough that they can provide sufficient raw material for genetic novelty.  Estimates of how often a given gene duplicates in evolution run about one duplication event per 100 million gene copies.  That seems low, but remember that there are thousands of genes in the genome, and, in many species (including Drosophila and now ours), there are hundreds of millions of individuals.  That means that, in the species, there are many genes that duplicate each generation.  Even if only a few percent of these survive inactivation, that’s a lot of raw material for evolutionary change.

• The presence of frequent gene duplications is supported by an independent study:  Emerson et al. (2008) found that in only fifteen lines of D. melanogaster from nature there were several hundred duplicate genes segregating as polymorphisms (that is, some individuals had one copy of a gene, some had two or more).  They estimated that 2% of the genome was tied up in this copy-number variation.  Clearly, there are a lot of duplicate genes variants floating around in nature.

The data of Chen et al., then, show that new genetic information can arise quickly, at least on an evolutionary timescale, and that the new genes rapidly assume new functions.  (Note: I am using Behe’s characterization of “new genetic information” as involving only new FCTs.  I don’t agree with this, since new genetic information can also arise when a single gene copy changes sufficiently to do something new.)

Although this doesn’t answer the question of what proportion of new evolutionary traits involve changes in gene sequence versus changes in gene regulation, it does show that a substantial part of the genome in one group of eukaryotes arises by the evolution of new FCTs that become involved in new developmental networks.  In other words, Behe’s conclusion from short-term lab studies of bacteria and viruses doesn’t apply to this well-studied group of organisms—and probably not to other eukaryotes, either.  All the evidence tells us that a rapid and important way to create new genetic information is through the duplication of genes and then their divergence by natural selection.

Poems are made by fools like me
But only selection makes an FCT

Now ID advocates like Behe could—and do—suggest that maybe the successfully duplicated-and-diverged genes didn’t arise by natural selection, but appeared by the instantaneous intervention of the designer (aka God/Jebus). But that idea is nixed by at least two observations.  The first is the appearance in many groups of dead, nonfunctional pseudogenes that were unsuccessful duplicates.  If the Great Designer made gene duplications to create genetic novelty, he surely failed in the majority of cases, and left his failures sitting around in the genome.

The second is the correlation between the age of a new gene and the type of selection acting on it.  If a Great Designer created these duplicates de novo to have a new function—presumably because natural selection couldn’t take a gene to a new function by gradual stepwise evolution—they would show instantaneous changes of DNA sequence that looked like selection, and then an instantaneous cessation of that selection right after the gene got its newly created function.  But that’s not what we see. What we see is not instantaneous but gradual change:  the younger a gene is (as estimated by the position on the evolutionary tree where it arose), the more rapid natural selection acts.  That directional selection continues to act as the gene gets older, but then slows down and finally becomes purifying selection, so that new DNA changes are eliminated.  This pattern is precisely what’s predicted if duplicates arise by accident and then quickly change by selection to assume new functions.

I suppose Behe and his minions will find a way to explain these two patterns by intelligent design, but that’s because ID theory isn’t science: there is no conceivable observation that can prove it wrong.  Every bit of data, no matter what it is, can always be fitted into the ID scheme, especially since its advocates allow a little bit of Darwinian evolution and posit an unpredictable and unknowable Designer.  But let us not tarnish the nice results of Chen et al. by using them to cast aspersions on ID.  They are a valuable contribution to the real science of evolutionary biology, showing how fast new genetic information can arise by gene duplication.

h/t: Manyuan Long for his patient explanations.

______

Chen, S., E. Zhang, and M. Long.  2010.  New genes in Drosophila quickly become essential. Science 330:1682-1685.

Emerson, J. J., M. Cardoso-Moreira, J. O. Borevitz, and M. Long. 2008. Natural selection shapes genome-wide patterns of copy-number polymorphism in Drosophila melanogaster. Science 320:1629-1631.

A new Gallup poll released three days ago confirms that America is still stuck on creationism, and that the numbers have barely budged in thirty years.  Here’s the upshot in a graph (click to enlarge):

Four in ten Americans are still straight-out creationists; those numbers have been pretty constant since 1982 though there’s a drop of four points since the last survey.  Of the 54% of Americans who accept some form of evolutionary change,  7 in 10 (a total of 38%), are theistic evolutionists, believing that God guided the evolutionary process.  Need I add that this form of evolution is not what we biologists accept?  And a total of 16% of Americans accept real evolution—purposeless and unguided by God.  That’s up from 9% in 1982, and may be a real trend.  Still, it’s dispiriting to realize that fewer than one in six Americans accepts evolution in the way scientists accept it.

As always, acceptance of evolution is positively correlated with level of education; here’s the table from the Gallup survey.  Still, only one in four Americans with a postgraduate degree accepts real, unguided evolution:

And, again, acceptance of evolution is highly correlated with religiosity. Churchgoers are nearly three times more likely to be straight-out creationists than those who don’t go to church, and only one-fifteenth as likely to accept scientific, unguided evolution.  Surprisingly, there’s not much difference between churchgoers and heathens in their acceptance of theistic evolution:

More non-news:  Republicans—and this surely reflects in part their greater religiosity—are more creationist than Democrats, and only 40% as likely to accept unguided evolution.  Dems and independents are about the same.

If there’s any good news here—and there’s not much, since the data still show us to be a benighted, science-rejecting people—it’s the increase in non-theistic evolutionists in the last decade.  I take no heart in the higher proportion of theistic evolutionists, since these range all the way from “God-started-the-process-and-let-it-goers”, to ID advocates who posit continual interjections of God into nature, to old-earth creationists who accept microevolution but not macroevolution, to people like Simon Conway Morris who see God as having intervened (or designed the evolutionary process) to permit the evolution of servile, God-worshipping humans.

What do we do?  My solution has always been to loosen the grip of religion on America, since that’s the overwhelming source of creationism.  Or maybe we just need some Rock Stars of Evolution.  Put me next to Lady Gaga, or Dawkins next to Katy Perry, and Americans will become Darwinists!

___

UPDATE:  Several people have touched on the issue of the relationship between Gnu Atheism and the putative increase of evolution acceptance.  If you take the evolution-acceptance trend as real, the rise began about 2001, and that was before the first Gnu Atheist book, Sam Harris’s The End of Faith, was published (that was 2004).  Still, what one can say is that since accommodationism by the NCSE and other organizations has been pretty constant over several decades, the rise of Gnu Atheism certainly hasn’t hurt acceptance of evolution—as many accommodationists maintain.  (Remember, though, that the “trend” may be specious.)

My colleague Jim Bull at the University of Texas at Austin conducted several of the phage experiments that creationist Michael Behe mentions in his recent Quarterly Review of Biology paper on experimental evolution.  As you may recall, Behe reviewed short-term studies of adaptive evolution of bacteria and viruses in the laboratory, and concluded that nearly all of this adaptation resulted from either inactivation of genes or the accumulation of small changes in genes (e.g. single nucleotide substitutions in the DNA) that caused a quantitative but not qualitative change in gene activity.

Behe’s implicit conclusion was that evolution in nature—and not just in bacteria and viruses, but all species—also occurred in this way; that is, brand-new genes or genetic elements (he calls them “FCTs”) could not originate de novo by mutation and natural selection, but had to be put there by the Intelligent Designer (aka God/Jebus).  Behe did not, and could not, say this in the paper, but intelligent-design advocates certainly touted this conclusion (see here and here, for instance), and now Behe himself has said the same thing on his blog at Uncommon Descent:

. . . I was saying that, no matter what causes gain-of-FCT events to sporadically arise in nature (and I of course think the more complex ones likely resulted from deliberate intelligent design http://tinyurl.com/32n64xl), short-term Darwinian evolution will be dominated by loss-of-FCT, which is itself an important, basic fact about the tempo of evolution.

Note that here he doesn‘t limit this conclusion (which he conveniently omitted from the QRB paper) to bacteria and viruses.

I’ve criticized Behe’s paper because in nearly all the laboratory experiments the researchers deliberately left out an important source of new genes and genetic elements that applies to bacteria and viruses in nature: the uptake of DNA (via “horizontal transmission”) from other species.  The experiments reviewed by Behe, then, weren’t a good model of what would happen to evolving microbes in nature, which are in fact often known to absorb new genes from distantly-related species. Further, in eukaryotes (organisms with “true cells”) we know that evolution has involved the creation of new genes and gene-controlling regions via gene duplication and divergence. (I’ll post more on this later today).  Thus, while Behe’s conclusions are valid for his particular, limited group of laboratory studies, they simply can’t be extended willy-nilly to evolution in nature.

I sent Behe’s paper to Jim Bull (who had already read it in draft) along with my replies, and asked him to comment.  Here, for the record, is his response.  This is somewhat technical, but will be useful to those who have read Behe’s paper, and I thought it deserved to be posted for the scientific record.  (Terminology: “phages” are short for “bacteriophages,” and refer to those viruses that infect bacteria.)

I’ve been asked to comment here on the paper by Michael Behe (MB), in large part because it reviews work of mine and because it is the focus of some controversy.  I read the paper in draft form some months ago and have not re-read it, but even then it exhibited an impressive command of the experimental evolution literature, at least the literature on adaptation of whole genomes of bacteria and phages (as opposed to the ‘directed’ evolution of genes on plasmids and of naked nucleic acids).

I consider MB’s characterization of most molecular evolution in these experiments as point mutations and/or deletions to be accurate.  Indeed, as I told MB in my comments on his ms, I had made the same point in a recent book chapter.  We have not seen the evolution of much novelty in these lab experiments on bacteria and viruses, at least not the classic gene duplication followed by diversification into new functions.  There is, however, a literature on what is known as directed evolution that does document the evolution of novelty when strong selection is applied to large populations, but those studies focus on individual genes (e.g., on plasmids) or short nucleic acids (e.g., 50-base RNA molecules).

What surprises me is that anyone would consider this absence of novelty in experimental evolution studies to be surprising, given what we know both about evolution and about the nature of the experiments.  As Jerry Coyne (JC) commented recently, the organisms and conditions used for those studies are not amenable to many of the types of evolutionary mechanisms and selective conditions that we think operate in nature.  The natural environment for many microbes includes lots of  free ‘environmental’ DNA from many sources, produced when cells die and release their DNA.  In addition, phages abound in natural environments, providing a ready means of DNA transfer between different bacteria, but many bacteria are also capable of incorporating environmental DNA into their genomes.

Finally, selection is not spatially uniform in nature, so that different individuals of the same population can be exposed to a wide variety of selective conditions.  In contrast, Rich Lenski’s famous lines use a single strain of E. coli in minimal media, transfered daily to fresh media.  No phage or other species of bacteria are present.  That E. coli strain is not capable of taking up DNA from the environment (which is irrelevant in these experiments anyway, since the only DNA in that environment would be from other E, coli of the same population).

Likewise, the phages used for most of the experimental work do not easily incorporate foreign DNA.  (MB used one of my studies with T7 to criticize JC for making this claim, but even here there’s a problem: what we failed to point out in our paper, and is fatal to MB’s criticism, is the fact that T7 degrades E. coli DNA, so even if the phage did incorporate an E coli gene, it might well destroy itself in the next infection.)  Other phages used in experimental evolution studies likewise are not known to incoeporate or tolerate host DNA or RNA, with the exception of M13.

Some experiments have attempted to select new functions and shown limited success.  One by Ichiro Matsumura selected the E. coli gene encoding beta glucoronidase to degrade a different substrate (a beta galactoside).  With strong selection over several cycles, he obtained a 500-fold gain in enzyme activity on the new substrate, but it was still not enough activity to allow the bacterium to grow on that substrate.  Four point mutations were responsible for the change in activity.  This study (and there are no doubt many parallels in the directed evolution  literature) suggests that evolution of new function from an old function likely is a slow process under more natural circumstances and requires a delicate combination of environmental conditions affecting selection.  Matsumura’s result certainly supports the possibility of evolution of new function from old function, but the design did not allow divergence between two identical copies—only one copy was present in the cell.

A study by Alisha Holloway, Tim Palzkill and me attempted to select the divergence between two copies of a bacterial gene.  Two identical copies of the gene for degrading ampicillin – Bla (for beta lactamase) – were put into the same cell, each on their own plasmid, and divergence was favored by growing the bacteria on two different antibiotics (ampicillin and a cephalosporin).  This seemed to be a sure-fire way to evolve divergent functions and maintain both copies, because of two observations already published.  First, although the initial Bla could not degrade the cephalosporin, it could evolve to resist the cephalosporin with one or two point mutations.   Second, resistance to the cephalosporing came at some cost in the ability of the mutated gene to resist ampicillin.  In principle, therefore, the only way for the cell to tolerate the two drugs was to maintain one copy of Bla unaltered and evolve the other copy to resist the cephalosporin.   Yet although cephalosporin resistance evolved in one of the two copies of Bla, the other copy of Bla was still lost — the cephalosporin resistant gene still had enough resistance to amp to allow the bacteria to grow.  This study merely illustrates that the conditions favoring the maintenance of two copies undergoing evolutionary divergence are delicate.  If we had 30 examples of these kinds of studies, and all gave negative results, we might begin to question the ‘duplication-then-divergence’ model.  But we already have plenty of evidence that new functions can evolve from old (e.g., the Matsumura study, and even the evolution of cephalosporin resistance from Bla); where we don’t have much effort yet is laboratory studies of divergence between two copies of what used to be the same gene.

My own view of the MB paper is that it has done a service to the study of evolution by pointing out where the next generation of experients should focus.  We don’t yet have many studies on the long term evolution of protein novelty (to get extreme divergence), and the types of selection used are typically extreme.  Answers to these problems aren’t yet available simply because we simply have not applied much effort.  Indeed, there is still much we have yet to understand about the seemingly more mundane process of point mutation evolution in the simplest environments.  As I noted above, we do have many dozens of ‘directed evolution” studies in which various functions and activities haven been evolved from random libraries of RNA molecules, and those studies have shown that selection can be a powerful and creative force.

A h/t to Dr. Bull for providing this response.  Here’s a photo of him guarding his property, the Double Helix Ranch in Texas, where he and co-owner David Hillis breed longhorn cattle: