Although I haven’t read much by David Dobbs, I’m told he’s a good science writer. But you couldn’t prove that from his latest effort in Aeon magazine: “Die, selfish gene, die” (the subtitle is “The selfish gene is one of the most successful science metaphors ever invented. Unfortunately, it’s wrong.”) I was going to write a critique in a single post, but I see that it would be too long, so I’ll divide it into two parts.
When I read Dobb’s title and subtitle, I thought “Whoa!” I keep up with evolutionary biology, and I wasn’t aware of any serious objections to the selfish gene metaphor. (Granted, there are misguided folks like Mary Midgley who don’t understand the metaphor, but they don’t count.) After all, the metaphor simply means that, during the process of natural selection, genes “act” as if they were selfish. And that means that those genes that replicate faster than others—those that make their “vehicles” leave more copies of those genes—spread through the gene pool, out-competing other gene copies. It’s a metaphorical and, to me, enlightening description of natural selection, for it helps one see more clearly how evolution works. For example, “meiotically-driven” genes, which don’t really improve the “adaptations” of their carrier but act by simply killing off the other gene copy in the gonads (each gene is present in two copies), are “selfish” but not conventionally “adaptive.” And they spread exactly as if they were selfish. This shows one difference between natural selection and adaptation.
At any rate, Dobb’s goal is several fold. First, he wants to claim that the metaphor of the selfish gene is wrong. Second, he wants to show that it’s wrong because new understanding of gene regulation—how genes turn on and off during development—render the selfish gene metaphor passé. Finally, he claims that a new theory, that of “genetic accommodation,” relegates much of conventional evolutionary theory to the dustbin, for the new theory deposes the centrality of the gene in favor of the centrality of the environment and its non-genetic effects on development. I’ll deal with the first two issues today, and the third tomorrow.
All three of these claims are wrong. However good Dobbs’s other writing may have been, this is a dire piece: one that is loaded with misinformation, wrong information, misleading information, and unsupported speculation. None of it even comes close to deposing the value of the “selfish gene” metaphor, and I’m not saying that just because I’m friends with Richard Dawkins. Indeed, it’s with a heavy heart that I set out to write this post, because I’ll have to devote at least 4 hours over two days to pointing out what’s wrong with Dobb’s article, and most people won’t read this anyway. I’m writing this only to set the record straight, to give my response to the several readers who sent me this article while praising it, as well as to those who have touted it on the internet as a wonderful and informative piece, and, finally, in hope that those interested in evolution might learn how to distinguish the wheat from the chaff in popular science writing. Sadly, even some science writers have praised it:
Dobbs begins with a gratuitous slur (disguised as praise) about Richard Dawkins:
These days, Dawkins makes the news so often for buffoonery that some might wonder how he ever became so celebrated. The Selfish Gene is how.
That comment is completely out of line in a serious popular article. Everyone who knows about evolution knows how Dawkins became celebrated: by writing and talking first about evolution, and then, more recently, about atheism. He succeeded brilliantly at both. But Dobbs can’t refrain from making fun of what, I suppose, are Dawkins’s statements on Twitter.
And then, after that faint praise, Dobbs tries to demolish the gene-centric view of the selfish genes. I’ll put in bold Dobb’s claims that I contest (indented quotes are from his piece):
New discoveries in gene regulation make the gene-centric view outmoded. Dobbs seems to conflate two issues and, on top of that, makes an error about the evolution of gene regulation. He begins with an example of what we call “polyphenism”: the change in an organism’s form over its life cycle. Dobbs’s example is that of grasshoppers turning into locusts (they’re the same species): when certain species of grasshoppers are crowded, they undergo striking changes in form and color, turning into the migratory insect that wreaks so much damage on crops. Other, uncrowded populations remain grasshopper-like.
Another example of polyphenism are castes of hymenopterans (mentioned by Dobbs). All female bees in a nest share similar genomes, but those fed royal jelly as larvae (an environmental effect) turn into queens, for that feeding activates different genes. The different cases of worker ants in some species, with some being soldiers and others trash-disposers, are a further example.
Similar striking changes in form occur also occur during the development of many species: male deer grow antlers, and, more dramatically, some species experience huge changes in form during their life cycle, as when a caterpillar undergoes metamorphosis into a butterfly.
Now these are all clearly examples of changes in gene expression: when a caterpillar turns into a butterfly, or a deer grows antlers when it ages, that must be due to differential gene expression—some genes are turned on and others off. The same holds when crowded grasshoppers become locusts or bee larvae become queens. This must be the case, for an individual has the same set of genes over its life, and so if it has programmed changes in behavior, morphology, or physiology during its lifetime, those changes must be due to the differential expression of the set of genes it inherits as a zygote. (The development of different cells and organs in the body reflects the same thing: they all have the same genome, but a liver cell differs from a heart or brain cell because it’s expressing a different set of genes.)
The changes above are all adaptive ones—grasshoppers are better off migrating when they’re crowded because they’ll run out of food if they stay put, and it’s obviously good, as the Tin Man found, to have a heart as well as a brain. Ergo, the changes in gene regulation that cause changes in form over one’s lifetime are evolved changes, created by the process of natural selection. In other words. the bits of DNA determining whether genes turn off and on (the so called “regulatory genes” or “regulatory elements”) have evolved. And they’ve evolved as selfish replicators. Bits of DNA that produce adaptive responses to environmental changes, like those that turn sedentary grasshoppers into migratory locusts, will be favored by natural selection and sweep through the species. They are no different from any other “selfish” gene that promotes the reproduction of itself or its carrier.
Note that this kind of evolution is precisely that outlined by Dawkins in The Selfish Gene. The difference between migratory grasshoppers and species that never transform is an evolved difference, and presumably an adaptive one. There are no principles here different from that those outlined by Dawkins. Adaptive differences between species that depend on differences in gene expression rather than differences in DNA sequences of structural genes (e.g., the hemoglobin gene) are both based on evolved differences in the DNA. The only difference is that the DNA changes reside in regulatory elements rather than full genes that code for proteins.
But Dobbs implies that somehow evolved differences in gene regulation do not involve “selfish-gene-ology” (the wrong or speculative parts are bolded):
Transforming the hopper is gene expression — a change in how the hopper’s genes are ‘expressed’, or read out. Gene expression is what makes a gene meaningful, and it’s vital for distinguishing one species from another. We humans, for instance, share more than half our genomes with flatworms; about 60 per cent with fruit flies and chickens; 80 per cent with cows; and 99 per cent with chimps. Those genetic distinctions aren’t enough to create all our differences from those animals — what biologists call our particular phenotype, which is essentially the recognisable thing a genotype builds. This means that we are human, rather than wormlike, flylike, chickenlike, feline, bovine, or excessively simian, less because we carry different genes from those other species than because our cells read differently our remarkably similar genomes as we develop from zygote to adult. The writing varies — but hardly as much as the reading.
This raises a question: if merely reading a genome differently can change organisms so wildly, why bother rewriting the genome to evolve? How vital, really, are actual changes in the genetic code? Do we even need DNA changes to adapt to new environments? Is the importance of the gene as the driver of evolution being overplayed?
Sorry, but he’s wrong here. First. he implies that, within a species, adaptive changes that evolve by natural selection don’t involve “genetic distinctions”: changes in DNA sequence. That’s wrong; in fact, it ‘s not even wrong! Even if species differences rest on changes in how genes are “read,” those differences in “reading” must themselves reside in the DNA, and thus must evolve as all selfish replicators evolve. Our differences from cows and cats do indeed reside in differences in the genome between species, but those differences can involve both the DNA sequences coding for proteins as well as in the sequences of “regulatory elements”. All heritable differences between species, in fact, must reside in the DNA; we know of no cases in which they don’t. Where else could they be?
Second, Dobbs conflates changes within an individual’s lifetime—changes that must be due to differential expression of genes carried by a single individual—with differences among species. The former must involve evolved differences in gene expression, while the latter can involve changes in both the sequences of proteins (“structural changes”) or gene expression (“regulatory changes”). There is an active controversy in the field—a controversy in which I’ve been involved—over the proportion of adaptive differences between species that involve changes in gene regulation versus those involving changes in the sequence of genes that make proteins (and hence the sequence of proteins). We know examples of both types of evolutionary change: genes that differentiate marine versus freshwater forms of stickleback fish, for example, seem to involve changes in gene expression, while the production of “antifreeze proteins” in Antarctic fish, differences between hemoglobins of birds that fly at high vs. low altitude, or insecticide resistance in insects, involve changes in protein sequences. But in all cases the evolutionary changes came from natural selection acting on selfish replicators. So we need not throw out the notion of “selfish genes” here, so long as we keep in mind that some genes, or bits of DNA, control the expression of others.
Differences between humans and other species cannot reside in differences in their DNA sequences. Dobbs makes another mistake: he argues that humans’ 20% DNA difference from cows is not enough to explain our profound morphological and behavioral differences. But it must! What other explanation is there? In fact, a 20% difference in DNA sequence adds up to a lot of genetic differences when you consider that the human genome contains 3 billion base pairs. A 20% difference is 600 million differences! Now many of the nucleotides reside in junk DNA, but if you consider that the human genome still has 21,000 genes, and assume (I’m just floating reasonable estimates) that genes have an average size of 800 DNA base pairs, then the average number of nucleotide differences between the genomes of humans and cows is 3,360,000 (21,000 X 800 X 0.2). Even the oft-cited 1% difference between humans and chimps works out to 168,000 differences in the genes. As I noted, some of these differences reside in non-coding parts of genes (like junk DNA or third positions of triplet “codons”), but there are also other bits of DNA, not in these 21,000 genes, that can affect gene expression. So even between humans and chimps there are thousands of genetic differences, and between humans and cows there are millions! How can Dobbs confidently conclude that our 20% DNA difference from cows isn’t enough to explain why we don’t eat grass or walk on four legs? He can’t. He’s just trying to flummox us with misleading “similarity indices.”
So when Dobbs says “you are 80 percent cow,” don’t be misled.
When you consider that adaptive changes within a lineage, or adaptive differences between species, must involve natural selection acting on DNA, whether that DNA involves making proteins or regulating whether protein-coding genes are turned on or off, you have to ask where the new paradigm really is. Is there a there there? Dobbs quotes my colleague Greg Wray, who answers “yes”:
Gregory Wray, a biologist at Duke University in North Carolina who studies fruit flies, sees this flexibility of genomic interpretation as a short path to adaptive flexibility. When one game plan written in the book can’t provide enough flexibility, fast changes in gene expression — a change in the book’s reading — can provide another plan that better matches the prevailing environment.
‘Different groups of animals succeed for different reasons,’ says Wray. ‘Primates, including humans, have succeeded because they’re especially flexible. You could even say flexibility is the essence of being a primate.’
According to Wray, [Mary Jane] West-Eberhard and many others, this recognition of gene expression’s power requires that we rethink how we view genes and evolution. For a century, the primary account of evolution has emphasised the gene’s role as architect: a gene creates a trait that either proves advantageous or not, and is thus selected for, changing a species for the better, or not. Thus, a genetic blueprint creates traits and drives evolution.
This gene-centric view, as it is known, is the one you learnt in high school. It’s the one you hear or read of in almost every popular account of how genes create traits and drive evolution. It comes from Gregor Mendel and the work he did with peas in the 1860s. Since then, and especially over the past 50 years, this notion has assumed the weight, solidity, and rootedness of an immovable object.
But a number of biologists argue that we need to replace this gene-centric view with one that more heavily emphasises the role of gene expression — that we need to see the gene less as an architect and more as a member of a collaborative remodelling and maintenance crew. . .Wray, West-Eberhard and company want to depose genes likewise. They want to cast genes not as the instigators of change, but as agents that institutionalise change rising from more dispersed and fluid forces.
Do these people not realize that gene expression is itself caused by differences in DNA sequence, and that those regulatory bits of DNA evolve precisely like “structural” genes themselves? And why on earth must “flexibility” reside more on changes in gene expression than on the structure of proteins themselves? We have no idea whether changes in gene regulation evolve more rapidly than changes in gene structure. Finally, Dawkins himself emphasizes the interactive nature of genes in The Selfish Gene, using another metaphor that of genes as individuals rowing in a crew boat (the organism). They all have to stroke in harmony to make the boat move, but you can still make it move faster by substituting a stronger oarsman for a weaker one. Selection often proceeds by that route: replacing one gene form with another in a way that keeps the vehicle—the organism—moving forward.
Further, there’s no sharp distinction between “regulatory” and “structural” genetic elements. If you know anything about genetics, you’ll know that some bits of DNA that regulate others themselves produce proteins: the so-called “transcription factors”. And a change in a gene making a transcription-factor protein (the famous Hox genes are one example) can cause changes in how other genes are regulated. A mutated “regulatory” protein, for example, can change how it interacts with the genes it regulates, so that a structural gene has regulatory effects.
It is still an unsolved question what proportion of adaptive differences between species reside in the DNA that makes proteins versus in the non-protein-coding “regulatory elements” of DNA that turn structural genes on and off. (Those elements are often adjacent to the genes they regulate, but need not be, so finding how and where a gene is regulated is often a complicated issue.) But that doesn’t matter, for in every case bits of DNA that are responsible for adaptations evolve “selfishly”: bits that improve an individual’s reproductive output become overrepresented in the next generation.
Nor does it matter that polyphemism, like the production of locusts from crowded grasshoppers, rests—as it must—on differential gene expression. For the ability of a genome to respond adaptively to environmental change by changing the expression of its constituent genes is itself an evolved phenomenon. (Another example is how Daphnia grow fish-deterring spines when placed in ponds that contain fish.) And the only way that adaptive evolution can occur is through the “selfish” behavior of genes.
An historical error. Maybe this is nit-picking, but it shows the level of Dobbs’s scholarship about the history of genetics. Dobbs claims that the idea of “kin selection,” or “inclusive fitness” came from the three founders of mathematical population genetics: Ronald Fisher, J.B.S. Haldane, and Sewall Wright:
Fisher, Haldane and Wright, working the complicated maths of how multiple genes interacted through time in a large population, showed that significant evolutionary change revealed itself as many small changes yielded a large effect, just as a series of small nested equations within a long algebra equation could.
The second kink was tougher. If organisms prospered by out-competing others, why did humans and some other animals help one another? This might seem a non-mathy problem. Yet Fisher, Haldane and Wright solved it too with maths, devising formulas quantifying precisely how altruism could be selected for. Some animals act generously, they explained, because doing so can aid others, such as their children, parents, siblings, cousins, grandchildren, or tribal mates, who share or might share some of their genes. The closer the kin, the kinder the behaviour. Thus, as Haldane once said, ‘I would lay down my life for two brothers or eight cousins.’
The first part is right: those three men did reconcile Mendelian genetics with Darwinian evolution. But they did not solve the problem of altruism or kin selection with maths. True, Haldane’s quote, whose authenticity is dubious since it was reportedly uttered in a pub, suggested offhand that he knew relatedness was the key to the problem. (But even this is at issue: see the discussion on pp. 174-178 in Ullica Segerstrale’s new biography of Bill Hamilton, Nature’s Oracle.) But that is all Haldane supposedly said about the issues of altruism and kin selection. There were no “formulas quantifying how altruism could be selected for.”
The problem of altruism and selection through kin was not in fact solved until the 1960s, largely by Bill Hamilton, whom Dobbs mistakenly and snarkily describes as a “funny statistician with a shaggy haircut.” (Hamilton was not a statistician but both an accomplished naturalist and a theoretical biologist. He didn’t work on statistics itself, but, like all biologists, used them.) Others like John Maynard Smith also made contributions to the problem of animal cooperation. Anybody who has studied the history of evolutionary biology knows this, and I wish Dobbs had let someone who knew that history vet his piece. (Dobbs also gets the modern method of assessing gene expression wrong: we have moved beyond microarrays.)
Tomorrow I’ll deal with the other main point of Dobb’s article: that there is a new paradigm—the notion of “genetic accommodation”—that completely revises how we should view evolution. A preview: he’s wrong here, too, and on two counts. Dobbs mischaracterizes how “genetic accommodation” works, and he doesn’t mention that we have no credible examples of that process producing any adaptations. Genetic accommodation is a complicated and unparsimonious idea in search of data.
In contrast, we have plenty of examples of “regular” Darwinian evolution acting on standing genetic variation.
But that’s for tomorrow, as Professor Ceiling Cat has work to do.
UPDATE: Dobbs has, in the comments below, referred us to his website, where he clarifies what he really meant in this piece. I am not going to read that site until I finish critiquing the Aeon piece tomorrow, for I want to analyze the piece as it was published, without being influenced by Dobb’s clarifications.