Charney provided us with a comprehensive review of “neogenetic” phenomena – retrotransposons, copy number variations, mitochondrial DNA, aneuploidy, and epigenetics. It was a fascinating read and highlights the many exciting breakthroughs in our understanding of the human genome. Unfortunately, these scientific advances were used to argue that twin and family studies are therefore obsolete. This conclusion misses the mark, for a number of reasons.
-
1. A core premise of his argument was that the “neogenome behaves like neither category,” that is, neither genetic (G) nor environmental (E) parameter. Based on his descriptions, this does not appear to be true – instead, it behaves like both G and E. To the extent that monozygotic (MZ) twins are more phenotypically similar than are dizygotic (DZ) twins because of their neogenetic profiles, the neogenome will be absorbed into G. To the extent that MZ twins also differ phenotypically because of differences in their neogenetic profiles, the neogenome will be absorbed in E. The key here is that both can happen simultaneously (e.g., parenting has been shown to influence twin outcomes at genetic and environmental levels simultaneously; Burt et al. Reference Burt, Krueger, McGue and Iacono2003; Reference Burt, McGue, Krueger and Iacono2007). Loading on more than one component of variance in no way means that they are somehow omitted from heritability estimates – indeed, to the extent that they contribute to outcomes, neogenetic effects are already necessarily being included in the G and E estimates we obtain. This is not the first time our understanding of the specific effects within a component of variance has changed. The same thing happened with G × E (contrary to Charney's claim, G × E are not estimated in traditional heritability estimates, which in fact explicitly assume that there are no G × E) – we now know that G × E are included in our simple estimates of G and E (Purcell Reference Purcell2002).
-
2. Charney's gestalt argument was that Mendelian models of familial resemblance should be discarded in favor of Lamarckian models. Put differently, he argued that the similarity of sibling's DNA matters less than their neogenetic (including environmental) similarity. Although he offered circumstantial support for his hypothesis, he was unable to point to studies directly testing this proposition. Why is this? Because neogenetic effects are tissue specific, most cannot be examined until after death, a fact that makes research very difficult to do in humans. How should we resolve this? Are we doomed to an endless cycle of back-and-forths where we argue about whether genetic or neogenetic resemblance matters more? No, and it turns out that the empirical answer will likely be found in the “natural experiments” that Charney would have us discard. Twin and family studies make use of naturally varying degrees of genetic and environmental relationships to infer general patterns of etiology. What sorts of family studies might allow us to test Charney's overall hypothesis that genetics matter less than neogenetics? There are many possibilities. Half-siblings would be useful, for example, in that we could compare siblings who were reared together with a common mother (but different fathers; these siblings share their family environment, an average of 25% of their nuclear DNA, their mitochondrial DNA, and many aspects of their prenatal environment) to those with a common father (but different mothers; who share their family environment and 25% of their DNA but not their mitochondrial DNA or their prenatal environment). Should half-siblings who share a mother be more similar to one another than those who share a father, it would argue for the importance of the prenatal environment/mitochondrial DNA.
An even stronger design might seek to disaggregate genetic resemblance and (most) neogenetic resemblance via a sample of children adopted as embryos. Embryos created via IVF (in vitro fertilization) are three days old when they are frozen and so consist of only eight undifferentiated cells. One could thus compare siblings (conceived via IVF) born to their biological mother to biological siblings adopted as embryos and born to another family (all siblings would share their DNA and any neogenetic effects that preceded the third day of embryonic life, but only some would also share their prenatal and familial environments). The children adopted as embryos could be further compared to other siblings adopted as embryos, but with different biological parents (in this case, the siblings would share only prenatal and familial environments). To the extent that biological siblings are more similar to each other when born to the same mother as when born to different mothers, it would highlight neogenetic/prenatal and environmental processes. To the extent that sibling similarity varied instead with degree of genetic relatedness, it would argue for the importance of DNA to sibling similarity.
Yet another design might take advantage of the fact that some MZ twins have different chorions, while others do not, thereby allowing researchers to examine the consequences of different degrees of prenatal sharing while controlling for the general effects of DNA. In fact, some such studies have already been done (Jacobs et al. Reference Jacobs, Van Gestel, Derom, Thiery, Vernon, Derom and Vlietinck2001; Reise Reference Reise1999) and have revealed minimal/inconsistent effects of chorion type on personality and cognitive ability. Jacobs et al. (Reference Jacobs, Van Gestel, Derom, Thiery, Vernon, Derom and Vlietinck2001), for example, found that whereas genetic effects accounted for more than 60% of the variance in cognitive abilities, chorion type accounted for less than 15% (and typically 0–2%). Although the above results begin to argue against prominent prenatal effects on normal-range outcomes (and moreover, suggest that the equal environments assumption need not be violated), future research should clearly seek to examine this possibility in more detail.
In closing, I would further note that discordant MZ twins may prove to be the most useful research design for examining neogenetic processes in humans, because controlling for the basic effects of DNA allows for particularly strong counterfactual inferences regarding neogenetic effects on outcome (stronger in fact than within a given person). This is particularly the case because, despite the exceptions noted by Charney, twins are comparable to singletons (Moilanen et al. Reference Moilanen, Linna, Ebeling, Kumpulainen, Tamminen, Piha and Almqvist1999; Robbers et al. Reference Robbers, Bartels, van Oort, van Beijsterveldt, van der Ende, Verhulst, Boomsma and Huizink2010; van den Oord et al. Reference van den Oord, Koot, Boomsma, Verhulst and Orlebeke1995) for the majority of psychological and behavioral outcomes. Future work should seek to do just this.
Charney provided us with a comprehensive review of “neogenetic” phenomena – retrotransposons, copy number variations, mitochondrial DNA, aneuploidy, and epigenetics. It was a fascinating read and highlights the many exciting breakthroughs in our understanding of the human genome. Unfortunately, these scientific advances were used to argue that twin and family studies are therefore obsolete. This conclusion misses the mark, for a number of reasons.
1. A core premise of his argument was that the “neogenome behaves like neither category,” that is, neither genetic (G) nor environmental (E) parameter. Based on his descriptions, this does not appear to be true – instead, it behaves like both G and E. To the extent that monozygotic (MZ) twins are more phenotypically similar than are dizygotic (DZ) twins because of their neogenetic profiles, the neogenome will be absorbed into G. To the extent that MZ twins also differ phenotypically because of differences in their neogenetic profiles, the neogenome will be absorbed in E. The key here is that both can happen simultaneously (e.g., parenting has been shown to influence twin outcomes at genetic and environmental levels simultaneously; Burt et al. Reference Burt, Krueger, McGue and Iacono2003; Reference Burt, McGue, Krueger and Iacono2007). Loading on more than one component of variance in no way means that they are somehow omitted from heritability estimates – indeed, to the extent that they contribute to outcomes, neogenetic effects are already necessarily being included in the G and E estimates we obtain. This is not the first time our understanding of the specific effects within a component of variance has changed. The same thing happened with G × E (contrary to Charney's claim, G × E are not estimated in traditional heritability estimates, which in fact explicitly assume that there are no G × E) – we now know that G × E are included in our simple estimates of G and E (Purcell Reference Purcell2002).
2. Charney's gestalt argument was that Mendelian models of familial resemblance should be discarded in favor of Lamarckian models. Put differently, he argued that the similarity of sibling's DNA matters less than their neogenetic (including environmental) similarity. Although he offered circumstantial support for his hypothesis, he was unable to point to studies directly testing this proposition. Why is this? Because neogenetic effects are tissue specific, most cannot be examined until after death, a fact that makes research very difficult to do in humans. How should we resolve this? Are we doomed to an endless cycle of back-and-forths where we argue about whether genetic or neogenetic resemblance matters more? No, and it turns out that the empirical answer will likely be found in the “natural experiments” that Charney would have us discard. Twin and family studies make use of naturally varying degrees of genetic and environmental relationships to infer general patterns of etiology. What sorts of family studies might allow us to test Charney's overall hypothesis that genetics matter less than neogenetics? There are many possibilities. Half-siblings would be useful, for example, in that we could compare siblings who were reared together with a common mother (but different fathers; these siblings share their family environment, an average of 25% of their nuclear DNA, their mitochondrial DNA, and many aspects of their prenatal environment) to those with a common father (but different mothers; who share their family environment and 25% of their DNA but not their mitochondrial DNA or their prenatal environment). Should half-siblings who share a mother be more similar to one another than those who share a father, it would argue for the importance of the prenatal environment/mitochondrial DNA.
An even stronger design might seek to disaggregate genetic resemblance and (most) neogenetic resemblance via a sample of children adopted as embryos. Embryos created via IVF (in vitro fertilization) are three days old when they are frozen and so consist of only eight undifferentiated cells. One could thus compare siblings (conceived via IVF) born to their biological mother to biological siblings adopted as embryos and born to another family (all siblings would share their DNA and any neogenetic effects that preceded the third day of embryonic life, but only some would also share their prenatal and familial environments). The children adopted as embryos could be further compared to other siblings adopted as embryos, but with different biological parents (in this case, the siblings would share only prenatal and familial environments). To the extent that biological siblings are more similar to each other when born to the same mother as when born to different mothers, it would highlight neogenetic/prenatal and environmental processes. To the extent that sibling similarity varied instead with degree of genetic relatedness, it would argue for the importance of DNA to sibling similarity.
Yet another design might take advantage of the fact that some MZ twins have different chorions, while others do not, thereby allowing researchers to examine the consequences of different degrees of prenatal sharing while controlling for the general effects of DNA. In fact, some such studies have already been done (Jacobs et al. Reference Jacobs, Van Gestel, Derom, Thiery, Vernon, Derom and Vlietinck2001; Reise Reference Reise1999) and have revealed minimal/inconsistent effects of chorion type on personality and cognitive ability. Jacobs et al. (Reference Jacobs, Van Gestel, Derom, Thiery, Vernon, Derom and Vlietinck2001), for example, found that whereas genetic effects accounted for more than 60% of the variance in cognitive abilities, chorion type accounted for less than 15% (and typically 0–2%). Although the above results begin to argue against prominent prenatal effects on normal-range outcomes (and moreover, suggest that the equal environments assumption need not be violated), future research should clearly seek to examine this possibility in more detail.
In closing, I would further note that discordant MZ twins may prove to be the most useful research design for examining neogenetic processes in humans, because controlling for the basic effects of DNA allows for particularly strong counterfactual inferences regarding neogenetic effects on outcome (stronger in fact than within a given person). This is particularly the case because, despite the exceptions noted by Charney, twins are comparable to singletons (Moilanen et al. Reference Moilanen, Linna, Ebeling, Kumpulainen, Tamminen, Piha and Almqvist1999; Robbers et al. Reference Robbers, Bartels, van Oort, van Beijsterveldt, van der Ende, Verhulst, Boomsma and Huizink2010; van den Oord et al. Reference van den Oord, Koot, Boomsma, Verhulst and Orlebeke1995) for the majority of psychological and behavioral outcomes. Future work should seek to do just this.