Extended Heredity: An interview with Russell Bonduriansky and Troy Day
31 July 2018
Extended Heredity: A New Understanding of Inheritance and Evolution is a fantastic new book by Russell Bonduriansky and Troy Day about the role of nongenetic inheritance in evolution. There are many similarities between the views presented in the book and the extended evolutionary synthesis but there are also differences. Kevin Laland identified some of these in his book review for Science and we wanted to take a deeper dive into some of the key issues. Katrina Falkenberg asked Russell and Troy some questions about the evolutionary significance of nongenetic inheritance, and asked Kevin to elaborate on some points he makes in his review.
Katrina (to Russell/Troy): Russell and Troy, your new book Extended Heredity is a wonderfully accessible and informative assessment of nongenetic inheritance (NGI) in evolution. Thank you for talking to us today. Could you start by telling us about the history of NGI and why you think it was side-lined?
Russell/Troy: To this day, students are taught that the debate between proponents of a purely genetic model of heredity and a model that allowed for other kinds of hereditary effects (such as the transmission of environmental influences across generations) was decisively settled well before the middle of the 20th century. But when we actually dug into the history, we found that the debate was not settled by empirical evidence. Rather, the evidence always suggested that there was more to heredity than Mendelian genes, and a large and growing number of more modern studies now clearly demonstrate that nongenetic forms of inheritance exist in many taxa.
So why was NGI side-lined historically? Our best guess is that this occurred for a combination of ideological, political and technical reasons. From an ideological standpoint, it’s noteworthy that the idea of “hard heredity” – that is, the belief that parents can pass on to their children only what they themselves had inherited from their own parents, and that hereditary factors cannot be altered by environment or experience – originated with Francis Galton. Galton is best known today as the originator of eugenics – the idea that problems like crime, alcoholism and poverty are a result of “bad heredity” and can only be remedied by eliminating the problem individuals from the breeding pool. For example, eugenicists delighted in pointing out that alcoholic parents often had children who became alcoholics as well, and argued that this was a result of bad heredity (“nature”) rather than environment and upbringing (“nurture”). This idea was extremely popular in the late 19th and early 20th centuries, perhaps because it allowed social elites to see themselves as intrinsically better and not just luckier than other people. The logic of eugenics breaks down if poverty or alcoholism can result from environment and upbringing rather than unalterable hereditary factors, and so support for eugenics also required a commitment to hard heredity. Thus, we suspect that belief in hard heredity initially piggybacked on support for eugenics. By the time eugenics fell out of fashion after WWII, the idea of hard heredity was already firmly engrained.
Politics and the cold war also played a role. In the 1930s and ‘40s, biology in the USSR was subverted by a scientifically illiterate opportunist named Trofim Lysenko. He gained Stalin’s trust, accused geneticists of sympathizing with eugenics and capitalism, and proceeded to crush genetics research at Soviet universities and institutes. In place of genetics, he promoted a hodgepodge of pseudo-scientific ideas based on the role of environment in heredity. Lysenko’s activities aroused horror and outrage in the West and, in the charged atmosphere of the cold war, anyone who questioned the purely genetic view of heredity became suspect.
Finally, many of the cellular mechanisms that bring about nongenetic inheritance through the germ-line (such as the epigenetic machinery that regulates gene expression) were only discovered with the growth of molecular biology in the late 20th century. Likewise, demonstrating the transmission of environmental effects across generations requires sophisticated experiments and statistical tools that only became possible in recent years. Thus, the gene-centric view of heredity triumphed at a time when biology was technically unprepared for the study of nongenetic inheritance.
Katrina (to Russell/Troy): What are the main lines of evidence (or strongest data) supporting the role of NGI in evolution?
Russell/Troy: Over the past 30 years or so, biologists have been discovering an ever-growing variety of cellular, physiological and behavioral mechanisms that can bring about the transmission of traits across generations independently of the transmission of genes. So, in our book, we argue that today the existence of nongenetic inheritance is no longer in question. We also believe that the importance of nongenetic inheritance in areas like human health is now established beyond a reasonable doubt. But the role of nongenetic inheritance in evolution is much more difficult to establish. What we have today is a lot of circumstantial evidence suggesting that nongenetic inheritance could influence the course of evolution in a variety of ways. We know that nongenetic inheritance occurs, that it can generate a great deal of heritable variation in phenotypic traits, and that these traits are often relevant for fitness. These are all the required ingredients for evolution by natural selection, and many theoretical analyses now also show that incorporating nongenetic inheritance into our models could help us to answer many evolutionary questions. What we’re still lacking are direct, experimental tests of the role of nongenetic inheritance in adaptive evolution, particularly in natural populations. Such tests are starting to be carried out so hopefully we will know much more in a few years’ time.
Nonetheless, we already have strong circumstantial evidence of the role of nongenetic inheritance in human evolution. The best-studied example is the evolution of lactase persistence in some human populations that domesticated cows and used raw milk in the adult diet. In such populations, individuals who could digest milk efficiently as adults could obtain a considerable nutritional benefit, so genetic mutations that caused the lactase enzyme (which facilitates the digestion of milk) to continue to be secreted throughout life were favored by natural selection. This is a case of gene-culture coevolution, where both a nongenetic (cultural) factor and a genetic factor interact to produce an evolutionary outcome that would otherwise be highly unlikely. But the role of gene-culture coevolution is likely limited to humans and perhaps a few other species of animals. We want to know whether nongenetic inheritance can also affect evolution in other organisms and in non-cultural contexts, and this will require creative experimental studies on organisms such as plants, nematode worms, insects, and small mammals.
Katrina (to Russell/Troy): Actually, we’ve already had a couple of nice blog posts touching on the possibility of gene-culture coevolution in animals (a recent one by Rose Thorogood and an earlier article by Andy Whiten and Kevin). What do you think are the strongest arguments against the importance of NGI in evolution and how do you evaluate these?
Russell/Troy: Some critics simply ask, Where is the evidence? They point out that direct, compelling demonstrations of an important evolutionary role are still lacking. We certainly agree with this point but as we note in our book, absence of evidence is not evidence of absence. And as we explained above, although we still lack clear, definitive examples, all of the necessary conditions for adaptive evolution via nongenetic inheritance have been documented. Thus, it seems reasonable to expect that NGI might well be important in evolution and so to keep an open mind until scientists have had a chance to examine the issue in greater depth. We simply don’t know enough at this point to dismiss nongenetic inheritance as a factor in evolution.
Another line of argument is that nongenetic factors are not well suited to an evolutionary role. In particular, critics argue that most nongenetic factors are too unstable and too limited in their range of variation to produce interesting evolutionary outcomes. By contrast, there’s a vast range of possible genetic variation, and selection on genes can produce almost any kind of organismal trait. Moreover, DNA is very stable and mutations are rare, so genes can persist for many generations. Our response, once again, is that it’s premature to dismiss the role of nongenetic factors in evolution. We don’t deny that DNA plays a special role in evolution, but we argue in our book that nongenetic factors could also play important and interesting evolutionary roles. Some nongenetic factors actually appear to be stable over many generations. But even highly transient nongenetic factors can alter the course of evolution by interacting with genes. So we believe that much more research is required to understand the role of nongenetic inheritance in evolution.
Katrina (to Kevin): Kevin, your Science review of Extended Heredity is very positive. What do you see as particular strengths of the book, in terms of convincing the evolutionary biology community of the importance of NGI?
Kevin: I do like the book. Obviously Russell and Troy have studied these issues for years, both experimentally and through mathematical models, and that track record lends their book a real authority. However, what makes it particularly valuable is its accessibility. Anyone who has tried to get their head around the almost bewildering richness of epigenetic inheritance mechanisms, not to mention the myriad of other forms of extended heredity, will know just how challenging this literature can be. Russell and Troy do a brilliant job at getting across the key ideas without swamping the reader in molecular or technical details. It is a beautifully written book. And, of course, for those of us with a particular interest in the evolutionary implications of extra-genetic inheritance, the fact that Russell and Troy are able to hone their arguments to reach an evolutionary biology community is really helpful. Another attractive aspect of their treatment is the use of a formal mathematical approach, specifically the Price equation, to demonstrate how extra-genetic inheritance can be incorporated into evolutionary theory. This treatment is very helpful, both as a means to illustrate the evolutionary significance of extra-genetic inheritance, and to illustrate how, in principle at least, it is relatively straightforward to incorporate extended heredity into evolutionary theory. That combination of accessibility and rigor is very compelling, and I expect the book to have a real impact.
Katrina (to Kevin): In your Science review, you presented two main criticisms of the book. First, there is the issue of whether variation could be biased towards the adaptive. Could you summarize how your position differs from that offered in Extended Heredity?
Kevin: Sure. Russell and Troy discuss a number of skeptical arguments concerning the significance of extended heredity, as well as some of the contention surrounding it. In general, I think they do a pretty good job, but I do take issue with their evaluation of the claim that novel heritable variation could be biased towards variants that are adaptive. They attribute this claim to one of our extended evolutionary synthesis papers,1 but it actually derives from Marc Kirschner and John Gerhart’s work on facilitated variation,2 and in fact is not specifically about extended heredity at all, but rather what we call ‘constructive development’.
These are complex issues, and we should not be surprised when ideas stemming from another field are misunderstood. Many researchers have seemingly jumped to the conclusion that facilitated phenotypic variation necessarily requires a process that biases mutation (or other novel heritable variation) towards high fitness solutions. That is not the case. The theory of facilitated variation is entirely consistent with the observation that most genetic mutations (and extra-genetic effects) are neutral or deleterious.3-5
The critical idea with ‘facilitated variation’ is that the developmental systems of organisms evolve the capability to channel random genetic changes (or novel environmental inputs) in phenotypic directions that are potentially useful.2 Kirschner and Gerhardt proposed their argument on the basis of extensive systems biology data. However, recently there have been some nice computational investigations of the evolution of facilitated variation, which demonstrate how it spontaneously emerges during evolution.4,6-8 For instance, a simulation analysis of Merav Parter and colleagues found that facilitated variation would evolve in a wide range of environmental conditions and patterns of variability.4 They show how organisms not only evolve adaptive responses to conditions experienced historically by their ancestors, but are also able to generalize to future environments, and show high adaptability to novel goals. Environments that change in a systematic fashion seem to promote facilitated variation and allow evolution to generalize to novel conditions. Similar conclusions have been reached by other computational studies.6-8
The point here is that, in the context of debates over extended heredity, the expectation that novel phenotypic variation will be biased towards variants that are adaptive is a red herring. This claim is based on an understanding of how developmental processes work, and not specifically a claim about extra-genetic inheritance. In an attempt to refute the claim, Russell and Troy present an argument disputing that extra-genetic inheritance can drive adaptive evolution without selection, and (with caveats concerning learning) I agree with that argument. I think selection is almost always important to the propagation of heritable variation. Russell and Troy also question whether extra-genetic inheritance generates adaptive variation more frequently than non-adaptive variation, and here again we broadly agree – it probably doesn’t, although learning is an exception. However, to me these arguments are not really evaluating the claim that novel phenotypic variation will be biased towards variants that are adaptive, but rather the alternative hypothesis that evolution could occur through adaptive mutation alone. The relevant question here should be whether developmental processes (including but not limited to extra-genetic inheritance) generate functional phenotypic variation more frequently than might otherwise be expected, and the aforementioned computational theory answers strongly in the affirmative.
Russell/Troy: We appreciate the opportunity to try and clarify these issues as well. We agree completely with Kevin that the existence (or not) of facilitated variation does not rely on the existence of extended forms of heredity. Facilitated variation and nongenetic inheritance are two distinct and logically separate possibilities. Nevertheless, these two topics are very often discussed together by proponents of an extended evolutionary synthesis (along with other ideas like niche construction). Therefore, we felt it was important to address the topic in our book because the two are often conflated. We believe that a great deal of the skepticism towards extended heredity results from a spill-over of skepticism towards an extended evolutionary synthesis. And one of the main objections to the latter is an objection to the possibility of directed variation. We too have objections to the possibility of directed variation and so we wanted to make this clear while still promoting further research into the potential significance of nongenetic forms of inheritance.
Our interpretation of the EES position is based on statements on the nature of heritable variation from EES papers. For example, according to Laland et al. (2015), it’s traditionally assumed that “heritable variation will be unbiased,” whereas the EES proposes that “heritable variation will be systematically biased towards variants that are adaptive….” In our book, we interpreted such statements as challenging a classic and fundamental assumption about the nature of mutational variance – namely, the assumption that the effects of new mutations are unbiased with respect to fitness.
To understand the classic assumption that mutations are unbiased with respect to fitness, it’s useful to recall an analogy suggested many decades ago by the influential evolutionary geneticist and statistician Ronald Fisher. Fisher argued that a mutation can be likened to a turn of the focus knob on a microscope that’s out of focus. In terms of this analogy, the unbiased-with-respect-to-fitness assumption simply means that the focus knob is turned each time in a haphazard direction, and is therefore equally likely to make the focus better or worse. (In this one-dimensional case, a small turn will improve focus ~50% of the time, but real organisms are multi-dimensional, so a random change in a random trait can be expected to reduce fitness >> 50% of the time.) By contrast, the EES position seems to be that the focus knob will be turned non-randomly, such that a given turn will improve focus > 50% of the time.
Why do we find the EES position (as we’ve interpreted it above) implausible? For the EES assumption to hold, organisms must somehow “know” how to adjust their traits so as to achieve improved performance in response to a challenge from the environment (that is, which way to turn the focus knob). We can imagine only two mechanisms that would enable organisms to do this. The first mechanism is brain-power: an intelligent being can analyse a problem, find a rational solution, and then transmit this solution on to its descendants through learning. This certainly happens sometimes, but this cannot be a general mechanism of adaptation because most organisms lack brains (we will return to this question below). The second and much more general mechanism is natural selection: if this particular challenge has already been encountered by previous generations, the lineage might have evolved a mechanism for adjusting the features of offspring so as to give them an advantage in dealing with that challenge, or similar challenges. For example, plants that experience heavy grazing by herbivores often produce offspring that secrete herbivore-repelling toxins. But, in organisms lacking sophisticated reasoning abilities, could such an adaptive response occur in response to a challenge that’s truly novel? For example, if the environment becomes polluted with industrial chemicals, would plants, insects or mice possess the capacity to diagnose the problem and adjust the physiology or behavior of their offspring so as to make them adapted to the chemicals? We have not yet seen any convincing explanation of how this could work.
The computational studies mentioned by Kevin are very interesting, but they seem (to us at least) to raise doubts about this possibility as well. It is true in these studies that the ability of organisms to generalize to future environments does sometimes evolve, but from our reading this typically happens only when there is enough overlap between the features of the ancestral and future environments. In other words, organisms might adaptively adjust the phenotype of their offspring to environmental conditions if these conditions are similar to those experienced in the evolutionary history of the population. Granted in some of these studies the way in which environments are similar can be very subtle and difficult to identify, but natural selection seems to generate organisms capable of discerning these features. Moreover, while the structure of the genome and the plasticity of development will certainly bias mutational effects towards certain phenotypic outcomes, and might make beneficial mutations more likely in general, we would still expect mutations to be unbiased with respect to the organism’s needs in the particular environment in which it finds itself.
Kevin: Well, sadly Russell and Troy are not alone in interpreting us in this way, so maybe there was something unclear about our writing. However, I can state categorically that the EES position (at least, as represented by my coauthors and I) is not to expect > 50% of novel mutations to be biased towards the organism’s needs. For us, a bias (towards something) implies greater than random expectation, not more often than not. As to our distinct readings of the computational studies, I’m optimistic that our differences could be resolved with more work designed to clarify in what senses past natural selection prepares developmental systems for novel challenges. Kirschner and Gerhart’s emphasis on ontogenetic exploratory and selective processes seems to me particularly relevant here.
Katrina (to Kevin): Your second criticism is of Russell and Troy’s treatment of social learning and culture. Where do you identify deficiencies, and how could these be resolved?
Kevin: Well first, let me qualify this criticism by acknowledging the scale of the challenge: the mechanisms of extended heredity are frighteningly diverse, and the literature spread across numerous fields, which makes it hard for any one individual to be an expert on every aspect. Russell and Troy deserve credit for covering so much. However, I did find the book’s treatment of animal learning disappointing. The claims that “only cognitively sophisticated animals” could learn adaptive solutions to novel circumstances, and that maladaptive behavior (such as birds learning to access poisoned bait) would spread just as readily as accessing a novel food, are badly out of touch with the animal learning literature. This is unfortunate, because they undermine their own case.
The whole function of associative learning is to allow animals to produce adaptive solutions to novel challenges, whilst extensive data reveal diverse mechanisms ensuring socially transmitted information is typically adaptive.9 For instance, contrary to the example they give, there are experimental data showing some birds produce disgust displays on consumption of toxins, the observation of which leads to reduced copying.10
Learned behavioral innovation is now extensively documented in animals, and typically leads to the production of adaptive phenotypic variants,11 many of which are subsequently propagated through social learning (e.g. the spread of lobtail feeding in humpback whales,12 or the exploitation of palm hearts by orangutans13). Much learning is reliant on phylogenetically ancient and widespread associative learning processes, such as classical and operant conditioning. This kind of learning can be applied in an extremely flexible and open-ended manner, including from heterospecifics, which means that animals are not restricted to learning only about environmental features previously encountered by the lineage (e.g. established predators or foods). Instead, animals can also learn about entirely novel stimuli or events, and devise appropriate responses to them (e.g. birds learn to evade a novel predator14). Via learning, animals can therefore generate adaptive responses to conditions without prior evolution of dedicated traits with suitable reaction norms. Copying in animals is typically non-random and strategic, with evidence that individuals often disproportionately copy successful and high-payoff behavior,15 thereby enhancing the spread of adaptive variants (e.g. some insects and birds are known to copy the nest-site decisions of successful conspecifics and heterospecifics16).
Social learning allows for the horizontal propagation of adaptive phenotypic variants amongst unrelated individuals in a population, often rapidly, such as in the social transmission of predator recognition in minnows,17 or socially learned mating preferences in grouse.18 The fact that, in such instances, behavior patterns are propagated amongst unrelated individuals clearly demonstrates that some novel heritable variants can spread without conventional natural selection operating through survival and reproduction. It seems to me a logical extension of the extended heredity stance that we embrace a broader definition of evolution as something like change in heritable variation over time. However, by that broader definition, social transmission is evolutionary change, a conclusion that many researchers will find unnerving. We need to recognize that an additional selective process is operating here, in which, through their learning and decision-making, individuals sort between alternative ideas and behavioral options. These decision-making processes have been fashioned by past natural selection to be broadly adaptive, but it does not follow that the details of what animals learn are prescribed by their genes. What particular foods a given ape eats, or what particular song a whale sings varies between populations depending on local tradition. Natural selection may sometimes fine-tune the perceptual and motivational systems of animals (e.g. making snake-shaped objects salient to monkeys19) but there is little evidence that it has greatly constrained general learning capabilities. This allows learning to introduce novelty into phenotype space, and perhaps trigger conventional adaptive evolution through genetic accommodation (e.g. killer whale dietary traditions are driving genetic divergence20).
Unfortunately, few evolutionary biologists appear to be familiar with the animal social learning literature, but to my mind these data only strengthen Russell and Troy’s primary arguments.
Russell/Troy: We mostly agree with Kevin on this. We agree that behavioral innovation and social learning are important to survival in animals with brains. We also agree with Kevin that “…an additional selective process is operating here, in which, through their learning and decision-making, individuals sort between alternative ideas and behavioral options.” In other words, cognitive solutions and innovations can sometimes represent a mechanism of adaptation in their own right. There can be no doubt that cognition has played a key role in human evolution, and it has probably played some role in the evolution of other mammals, birds, and perhaps some other animals as well. Cognition can certainly result in effects akin to a mutational bias for fitness, in that intelligent agents can find solutions to the challenges posed by their environment. To take an obvious example, humans can identify industrial toxins and other hazards and devise ways to shield themselves and their offspring from those risks. Perhaps monkeys and birds are capable of similar cognitive feats in some circumstances.
If there’s any disagreement here, it might be about how far cognitive innovations and social learning can drive adaptive evolution without the help of natural selection. On this question, we take a more cautious position than Kevin does. It’s extremely difficult to gauge the long-term consequences of a behavior, and even humans are notoriously bad at this. Smoking and other deleterious cultural traits often spread and persist in human populations, so it’s very likely that birds and other animals are susceptible to this as well. Presumably, disgust displays and the responses that they elicit both evolved via natural selection. At any rate, the potential for fitness-enhancing cognitive innovations in humans and non-humans is a fascinating question for empirical investigation.
We should also add that the relatively limited discussion devoted to cognition and learning in our book does not reflect our lack of appreciation for this fascinating topic. We simply had a very broad literature to cover, and wanted to emphasize the diversity of nongenetic mechanisms of inheritance.
Katrina: So let me see if I can sum up your respective positions. You agree that extended heredity is probably widespread and may be evolutionarily important, although more research is required to clarify exactly how it’s important. You also agree that extended heredity is logically distinct from facilitated variation. However, there is some disagreement on the role of ‘facilitated variation’. I think you all agree that evolutionary history, genetic architecture, and developmental plasticity affect phenotypic variation in ways that probably have very important consequences for the rate and direction of evolution. Where you do seem to disagree is on a more specific point – the issue of whether ‘facilitated variation’ actually challenges the classic assumption that the effects of new mutations are unbiased with respect to fitness.
Russell/Troy: We do agree on a number of points. However, we feel that some further clarification is needed on how Kevin and his coauthors understand the consequences of ‘facilitated variation’ in this specific (and very important) sense, and believe that it would be very useful to formulate the EES position in more precise terms that would make it possible to see exactly how the EES diverges from the conventional position.
Kevin: I think that’s right.
Russell/Troy: We’re happy for this opportunity to discuss these issues. Thanks to Kevin for inviting us to engage in this discussion, and to Katrina for facilitating our dialogue.
Katrina & Kevin: Our pleasure.
1. Laland KN et al. 2015. Proc R Soc B 282:20151019. 2. Kirschner MW & Gerhart JC. 2005. Yale UP. 3. Gerhart JC & Kirschner MW. 2007. PNAS 104:8582-8589. 4. Parter M, Kashtan N & Alon U. 2008. PLoS Comput Biol 4:e1000206. 5. Uller T, Moczek AP, Watson RA, Brakefield PM & Laland KN. 2018. Genetics 209:949-966. 6. Draghi J & Wagner GP. 2009. J Evol Biol 22:599-611. 7. Watson RA, Wagner GP, Pavlicev M, Weinreich DM & Mills R. 2014. Evolution 68:1124-1138. 8. Kouvaris K, Clune J, Kounios L, Brede M & Watson RA. 2017. PLoS Comput Biol 13:e1005358. 9. Hoppitt W & Laland KN. 2013. Princeton UP. 10. Mason R 1988. Lawrence Erlbaum Associates. 11. Reader SM & Laland KN. 2003. Oxford UP. 12. Allen J, Weinrich M, Hoppitt W & Rendell L. 2013. Science 340:485-488. 13. Russon AE. 2003. Oxford UP. 14. Thorogood R & Davies NB. 2012. Science 337:578-580. 15. Kendal RL et al. 2018. Trends Cog Sci 22:P651-665. 16. Seppänen JT, Forsman JT, Mönkkönen M, Krams I & Salmi T. 2011. Proc R Soc B 278:1736-1741. 17. Chivers DP & Smith RJF. 1995. Ethology 99:286-296. 18. Gibson RM, Bradbury JW & Vehrencamp SL. 1991. Behav Ecol 2:165-180. 19. Mineka S & Cook M. 1988. Lawrence Erlbaum Associates. 20. Hoelzel A & Moura A. 2016. Heredity 117:481-482.
Buy the book:
Extended Heredity: A New Understanding of Inheritance and Evolution by Russell Bonduriansky & Troy Day (Princeton University Press, 2018)
Read the Science review:
Beyond Epigenetics by Kevin Laland (Science, 2018 360:1408)