Through a review of molecular findings, Charney revisits the important topic of gene–environment interdependence in behavioral traits. While a need for the revision of some concepts, and perhaps tenets, of behavioral genetics was recently heralded by geneticists (Petronis Reference Petronis2010), developmentalists (Rutter Reference Rutter2012), and social scientists (Tabery Reference Tabery2007), earlier contributions identified genetic influences upon environmental measures (Plomin & Bergeman Reference Plomin and Bergeman1991), and long before the bioinformatics era, behavioral geneticists had hypothesized epigenetic mechanisms to account for deviations from linear-additive genetic effects (Gottesman & Shields Reference Gottesman and Shields1982). So the issue does not seem to be whether the arguments presented by Charney and several other before him should influence behavioral genetics' concepts and methods, but rather how, and in what specific circumstances, this should take place.

Below, while focusing on epigenetics and gene expression, I provide some examples of how predictions stemming from Charney's treatment of “neogenomics” are at odds with empirical findings, and I underline that while epigenetic and gene expression effects are complex, they are partially predictable by the degree of genetic relatedness between individuals. By these arguments, I wish to emphasize that neogenomic effects do not always need to have pervasive and relevant consequences on phenotypes, nor do they necessarily contradict “classic” behavioral genetics' tenets. Hence, their action can at least in part be conceptualized, modeled, and incorporated into classic methods (Battaglia Reference Battaglia2012).

First, the relationship between widespread epigenetic marks and genetic expression is still controversial. Zhou et al. (Reference Zhou, Yuan, Mash and Goldman2011) showed that H3K4me3 changes of histonic profiles in the hippocampus of substance-addicted humans do not predict gene expression at the individual gene promoter level. Instead, H3K4me3 changes appeared to highlight macroscopic alterations of histone structure probably induced by addiction. Therefore, while epigenetic effects are potentially important, the individual and specific impact on brain and behavior is neither well understood nor unambiguously linked to gene expression data.

Second, according to Charney, through the effects of neogenomic forces (e.g., transposons), the degree of genetic heterogeneity between monozygotic (MZ) cotwins is expected to increase in time, as they experience increasingly different environments and lifestyles. Contrary to this prediction, for several phenotypes (e.g., general cognitive abilities) the degree of similarity among MZ pairs increases in time, even among more elderly (age>80) individuals (McClearn et al. Reference McClearn, Johansson, Berg, Pedersen, Ahern, Petrill and Plomin1997; Plomin Reference Plomin1986).

Third, based on the premise that physical exercise promotes hippocampal neurogenesis in laboratory animals, Charney concludes that this same mechanism is transferrable and relevant to humans, with the expectation that exercise can impact causally upon mental health and increase twins' genetic heterogeneity. This is based on the intuition that, following differential physical exercise, MZ twins' brains will exhibit enhanced, retrotransposon-induced genetic heterogeneity. Experimental observations contradict these hypotheses. De Moor et al. (Reference De Moor, Boomsma, Stubbe, Willemsen and de Geus2008) data show that the cross-sectional bivariate genetic correlations between indexes of physical exercise and symptoms of anxiety/depression differ quite remarkably from the longitudinal correlations. Among the longitudinal bivariate correlations between exercise and anxiety/depression, the only significant were the genetic, not the environmental correlations. Even in the presence of biased assumptions of genetic–environmental monozygotic–dizygotic (MZ–DZ) correlations, these data are inconsistent with an augmentation of intrapair genetic heterogeneity for the traits under scrutiny. Moreover, a series of tests of the association between exercise and mental health did not support a causal link. Instead, De Moor et al.'s data were consistent with shared genetic factors having opposite effects upon exercise and anxiety/depression.

The above examples show that while the field of neogenomics is conceptually interesting, some predictions are hardly generalizable or are contradicted by findings. It is argued that the parsimonious, classical behavioral genetic models that rely on simpler (occasionally even simplistic) assumptions and theory often prove quite effective in explaining empirical data. In evaluating the trade-offs between conceptual and computational parsimony and explanatory power, it should be remembered that the Slatkin (Reference Slatkin2009) study, which included within the same model both epigenetic and genetic factors to estimate complex traits' susceptibility, suggested a role for epigenetic changes in explaining individual risk, but not necessarily heritability. Thus, the power of epigenetic changes to explain important issues such as the “missing heritability” in genome-wide association (GWA) studies is currently unknown (Bell & Spector Reference Bell and Spector2011).

So can epigenetic and gene expression effects, together with their inherent variability, be at least partially ascribed to generalizable concepts, and thus studied?

There is wide individual variability for gene expression, even when measured in the same cell types at the same developmental stage; some genes show little, and other genes conspicuous, variation of expression levels. Since in these experiments (Cheung & Spielman Reference Cheung and Spielman2009) the non-genetic sources of variation are maintained the same for all the genes, and the measured degree of variability is less pronounced among related – compared to unrelated – individuals, it can be argued that we are facing yet another measure of heritability here, with the measure of expression (e.g., mRNA) becoming the phenotype under scrutiny. Thus, gene expression is itself a heritable quantitative trait, and the statistical power to detect gene variants that affect gene expression depends on such heritability (Visscher et al. Reference Visscher, Hill and Wray2008). Analogously, DNA methylation profiles (Kaminsky et al. Reference Kaminsky, Tang, Wang, Ptak, Oh, Wong, Feldcamp, Virtanen, Halfvarson, Tysk, McRae, Visscher, Montgomery, Gottesman, Martin and Petronis2009) are significantly more similar among MZ than DZ twin pairs. Thus, methylation per se could become the object of modeling: The methylation status of an individual in a given genomic region can be conceived as the result of linear-additive genetic and environmental effects plus a latent Ep causal component, the latter term indicating heritable and stable DNA-sequence independent epigenetic factors (Bell & Spector Reference Bell and Spector2011). Likewise, one may devise the incorporation of epigenetic effects into structural equation modeling of twin data. Inevitably, in the light of high indetermination and lack of robust findings, fixing intrapair correlations for methylation with certainty would not be easy (Bell & Spector [2011] proposed 0.4(CI: 0.5, 1) and 0.2 (CI: 0.7, 1) for MZ and DZ pairs, respectively).

Although several technical challenges remain, adding epigenetic effects to the classic behavioral genetics arsenal is conceivable. This could open new and conceptually fascinating avenues; for instance, stochastic and environmentally induced epigenetic variation could explain a sizable proportion of the effects currently attributed to non-shared environment.