Transgenerational inheritance: Models and mechanisms of non–DNA sequence–based inheritance

See allHide authors and affiliations

Science  07 Oct 2016:
Vol. 354, Issue 6308, pp. 59-63
DOI: 10.1126/science.aaf4945


Heritability has traditionally been thought to be a characteristic feature of the genetic material of an organism—notably, its DNA. However, it is now clear that inheritance not based on DNA sequence exists in multiple organisms, with examples found in microbes, plants, and invertebrate and vertebrate animals. In mammals, the molecular mechanisms have been challenging to elucidate, in part due to difficulties in designing robust models and approaches. Here we review some of the evidence, concepts, and potential mechanisms of non–DNA sequence–based transgenerational inheritance. We highlight model systems and discuss whether phenotypes are replicated or reconstructed over successive generations, as well as whether mechanisms operate at transcriptional and/or posttranscriptional levels. Finally, we explore the short- and long-term implications of non–DNA sequence–based inheritance. Understanding the effects of non–DNA sequence–based mechanisms is key to a full appreciation of heritability in health and disease.

Advances in molecular biology in the second half of the 20th century firmly established DNA sequence as the molecular substrate of inheritance (1). DNA seems to satisfy the requirements of both Darwin and Wallace’s evolutionary theory and Mendel’s laws of inheritance, now unified in the “modern synthesis,” as put forward by Huxley (2). DNA provides a substrate for random variation—for example, through mutation during replication. In the view of modern synthesis, biological forms or phenotypes interact with the environment and are the subject of natural selection, but the heritable substance or genotype is not. This strict separation of genotype from phenotype also led to the rejection of the inheritance of acquired traits. This was first formalized by Weismann in the germ-plasm theory (3) and is often also referred to as the “Weismann barrier.” The idea of the Weismann barrier is that the information flow from genotype to phenotype is strictly irreversible. This places the germline on a pedestal, responsible for all inheritance devoid of influence from somatic cells. Today, genetics is usually and appropriately equated with DNA sequence–based mechanisms. Yet, it appears that biology is much richer: Many phenomena and mechanisms of nongenetic and/or non–DNA sequence–based inheritance have been described in a range of model organisms, challenging our perception of the well-established relationship between transmitted genotype and phenotype. How can we learn more about the mechanism and effects of this extended type of inheritance?

What is the evidence for nongenetic or non–DNA sequence–based inheritance? Look beyond humans!

Given that DNA is the major substrate of inheritance, convincing evidence for inheritance that is not based on DNA sequence often emerges where genomic DNA can easily be experimentally controlled. This is the case for species that are parthenogenetic, self-fertilizing (hermaphroditic), or isogenic, as is the case for many laboratory model organisms. Even when genetics can be controlled, other caveats need to be considered: These include changes in the environment such as parental nurturing (4); microbiome; and, in mammals, milk (5) or even cultural influences. For example, in mice, maternal nurturing can have profound effects on phenotype (6). Finally, a useful distinction is often made between intergenerational and transgenerational inheritance. In the former, the environment of the parent can directly affect germ cells of the offspring. For the latter, in the case of mammals, a true transgenerational effect can be defined if transmitted to the F3 and possibly future generations, arising from a germline in the F2 generation that has not been exposed (7, 8). Examples of non–DNA sequence–based inheritance and/or the inheritance of acquired traits have recently been reviewed elsewhere (711) and span many diverse organisms. For instance, in plants, a naturally occurring mutant of Linaria vulgaris (toad flax) that was originally described by Linnaeus more than 250 years ago and dramatically alters the flower shape (Fig. 1) was found to be due to the methylation state of the Lcyc gene and did not involve alteration of the DNA sequence (12). In yeast, prions are a natural and common mechanism for the inheritance of traits independent of DNA (13). In the ciliate Tetrahymena, swimming behaviors in isogenic populations are dependent on the environment and are heritable (14). In animals, such as the nematode Caenorhabditis elegans, the RNA interference (RNAi) pathway results in non–DNA sequence–based heritable changes (1517), genetic manipulation can result in multigenerational phenotypic variation that is independent of DNA sequence (1823), and environmental perturbation can result in heritable phenotypic changes (8, 2426). In mice, a number of different types of nongenetic intergenerational or transgenerational inheritance have been observed, as reviewed in detail elsewhere (8, 9, 11, 27, 28). For example, genomic imprinting (as discovered in mice by Surani and Solter) results in the expression of genes being dependent on their parental origin, a process regulated by epigenetic mechanisms (29). In addition, specific loci show variable and heritable nongenetic changes in expression, such as at the Agouti viable yellow (Avy) locus affecting coat color and metabolic outcome, caused by the insertion of a transposable element, an endogenous retrovirus, providing a variably DNA methylated ectopic promoter for agouti (30). A number of such heritable effects, including at Avy, can be modulated by environmental influences. When considering environmentally induced effects, a particular emphasis has been put on nutrition and stress as inducers of nongenetic effects. For example, parental diet can affect the phenotype of the offspring (27, 3134), as shown in one recent study exploring metabolic outcomes in both male and female mice born to parents that consumed a high-fat diet (33). Early life stress is another example for which several rodent models have been reported (3537). An emphasis on nutritional models in mice might be the consequence of evocative epidemiological studies in humans that suggest maternal and paternal inheritance of nutritional states (38, 39). Although in most of the examples mentioned above the mechanisms of inheritance are unlikely to be DNA sequence–based, with varying strength of evidence, the mode(s) of transmission of nongenetic effects remain to be discovered.

“…a complete understanding of non–DNA sequence–based heritable effects requires a number of components, and we do not currently have the complete picture for any natural example.”

“Many phenomena and mechanisms of nongenetic and/or non–DNA sequence–based inheritance have been described in a range of model organisms, challenging our perception of the well-established relationship between transmitted genotype and phenotype.”

Fig. 1 Epialleles.

Shown is an example of a natural non–DNA sequence–mediated multigenerational epigenetic inheritance in toad flax: breaking flower symmetry (12). The flower changes from bilateral symmetry (left) to radial symmetry (right) via a naturally occurring mutant. The molecular basis of this mutant is a metastable and heritable DNA methylation state of the Lcyc gene locus.


How to reconsider the concept of inheritance without the restriction of the DNA model—Replication or reconstruction?

The semiconservative mechanism of DNA replication (40) provides a clear paradigm of how genetic information is faithfully transmitted during each cell division in mitosis and meiosis. This paradigm is so powerful that great emphasis has been placed on replicative inheritance of other information. Due to the well-understood mechanisms associated with the propagation of epigenetic states such as DNA methylation, experiments analyzing epigenetic modifications to DNA and chromatin have proved popular in attempts to explain the heritable memory of environmental experience. In both cases, enzymes have been identified that can “read” a modification and replicate it locally on the newly synthesized strand (in the case of DNA) or can propagate it on newly assembled histones on chromatin (41). Both mechanisms fit an appropriately narrow mechanistic definition of “epigenetics” that is discussed in detail elsewhere (42).

Yet, as pointed out by Jablonka (43), information might not only be copied or replicated but may also be reconstructed in each generation (Fig. 2). Whether transgenerational mechanisms leading to adverse outcomes are replicated between generations or reconstructed in each generation (or a combination of both) has implications for our consideration of the mechanisms themselves, as well as for the design of experiments to determine them. The prospect of reconstruction is possible at the Avy locus, where epigenetic reprogramming has been shown to occur in the preimplantation embryo despite the nongenetic transmission of the phenotype across generations. Metabolic or cell signaling loops or noncoding RNA expression are examples that might also fit a reconstructive model of inheritance. It is important to consider that epigenetic marks such as DNA methylation or histone modifications might be reconstructed with RNAs, transcription factors, or even metabolic loops acting as intermediate carriers of information. Heritable epigenetic information, including DNA methylation or histone marks, might therefore not be present in all cells of an organism but might be reconstructed in a specific tissue or tissues only. As a consequence, careful consideration of the underlying mechanisms of extended inheritance is relevant when considering specific paradigms and vice versa.

Fig. 2 Replication versus reconstruction as mechanisms of inheritance.

As proposed by Jablonka and Raz (9, 43), inheritance might be replicative in nature and involve direct copying of DNA and epigenetic marks (A) or, alternatively, inherited information might be reconstructed during development (B). In the latter case, the epigenetic marks are not present in all cells at all times. Instead, epigenetic information is reconstructed from another substrate (green cheese, RNA, etc.): “The laws of genetics had never depended upon knowing what the genes were chemically and would hold true even if they were made of green cheese” [Ed Lewis (76)].

When one considers the basis for transgenerational inheritance of phenotypic traits, the first and most likely port of call will, and should, be a genetic one. DNA for the storage of heritable information is often thought to be exceptionally conservative. The replication error rate is small (~10−8 per base pair) and surprisingly similar across many organisms (44). However, the DNA mutation rate is not necessarily random throughout the genome (45), can depend on epigenetic state (46), and might be tunable in response to environmental stress, at least in some organisms (47). In particular, DNA methylation is a mutagen, with recent examples of its potential effect being observed in the increased incidence of C-to-T transitions found in cancer genomes (46). Once a genetic mechanism can be excluded and non–DNA sequence–based mechanisms inferred, such information might also reveal very distinct behaviors. Instead of being deterministic, this information might be metastable, stochastically switching between two (or more) states every few or many generations, or might exhibit parental-origin effects that differ based on whether it is transmitted via sperm or egg. A clear example of state-switching is found with metabolic loops in prokaryotes that have recently been characterized quantitatively (48). Any state-switching might be completely dependent, independent, or tunable through the environment. In the latter case, prolonged environmental exposure, perhaps even over several generations, might be required for state-switching. All of the above have been observed (9). When discussing the heritability of environmental information, or acquired traits, a controversy between Lamarckian and Darwinian models of heritability is often evoked, with the former having been largely dismissed. Alas, the history of the science of heredity is more complex; this has been discussed in detail elsewhere (49). For example, Darwin’s theory of pangenesis suggested the presence of gemmules, nongenetic factors that could be transmitted via the germline to influence the inheritance of acquired characteristics (50).

What is the known mechanistic toolkit beyond DNA sequence?

A non–DNA sequence–based mechanism of heritability requires a substrate that transmits the information from one generation to the next. There must also exist a mechanism for the transmitted information to be “read” or interpreted in the offspring to alter the phenotype. Finally, when considering environmental cues, the transmitted information should also be tunable in the parent. Thus, a complete understanding of non–DNA sequence–based heritable effects requires a number of components, and we do not currently have the complete picture for any natural example. Nevertheless, diverse organisms provide clues and candidate mechanisms that we will examine here with regards to mammals, including humans (Fig. 3).

Fig. 3 Interaction between epigenetics and genetics.

Genomes contain mobile elements (red) and genes (green). Cellular defense mechanisms such as piRNAs (blue) silence the expression of transcripts from mobile elements by modifying the epigenetic state (e.g., locus-specific histone modification, DNA methylation). Silencing of mobile elements that have inserted close to genes might affect gene expression locally (state 1). This polar effect on gene expression might be influenced through intrinsic and extrinsic factors (state 1 versus state 2).

In plants, stable replicative inheritance of DNA methylation marks throughout many generations is a well-characterized mechanism (see above and Fig. 1). The simplicity of this mechanism and the fact that the DNA methylation machinery is conserved in mammals (51) makes this an attractive mechanistic model. The DNA methylation state of genomic DNA in somatic tissues can be modulated by the intrinsic developmental environment and age in mice and humans (52, 53) and hence has the potential to be extrinsically and directly modulated. Still, germline and preimplantation reprogramming in mammals are formidable barriers to replicative inheritance: In both the female and male germlines, DNA methylation is largely eliminated from the genome and can be considered a molecular manifestation of the Weismann barrier (54). However, in mammals, germline DNA methylation erasure appears to be incomplete (55), with some nonrepetitive and particular classes of repeat sequences being reported as somewhat resistant to reprogramming. The idea of certain sequences that might be refractory to germline epigenetic reprogramming provides a compelling mechanism for the inheritance of modulated epigenetic states. A second wave of genome-wide reprogramming occurs in the mammalian preimplantation embryo. At this developmental stage, mechanisms exist to prevent the reprogramming of particular sequences, such as genomic imprints (29). Related mechanisms might also be operational at other sequences, including those targeting repressive epigenetic states to repetitive endogenous retroviral sequences during and after the reprogramming period. These epigenetic targeting approaches might provide a mechanism for reconstructing epigenetic states between one generation and another and may involve not only DNA methylation but also the targeting of repressive histone modifications (56).

Histones, which are closely associated with DNA in the histone octamer forming the nucleosome, are a suitable candidate substrate of information that can be propagated during DNA replication through local nucleosome retention. Analogous to DNA methylation, classes of “reader” and “writer” proteins that can maintain histone marks have been identified for a number of posttranslational modifications (41). Evidence from nematodes, flies, mice, and humans suggests that histone modifications might be involved in heritable effects (19, 20, 5759). However, whether histone marks are heritable in a replicative manner across generations in mammals is questionable because they are subject to erasure during germline reprogramming, as is the case for DNA methylation (54). Furthermore, in the male germline, histones associated with genomic DNA are largely replaced with protamines in sperm. Immediately upon fertilization, the protamines are removed and the paternal chromatin rapidly remodeled using maternally deposited stores of histones. However, the few modified histones in sperm that are carried through to the egg upon fertilization have recently been linked to altered development in the offspring, indicating that their correct modification is important for offspring development (60, 61). Hence, similar to DNA methylation, histone modifications might contribute a replicated mechanism of inheritance but might also be reconstructed upon remodeling of the parental genomes after fertilization. Therefore, environmental influences have the potential to compromise important normal developmental processes associated with the germline and early developmental dynamics of epigenetic states.

RNA-mediated mechanisms

Although DNA methylation and histone modification represent epigenetic mechanisms that are compelling heritable environmental mediators of nongenetic states, this is not necessarily the case for RNA. However, specific long and short noncoding RNAs are tightly linked to both mechanistic pathways and, hence, can contribute to a dialogue between environment and genome. This is particularly well understood for the small noncoding RNAs involved in RNAi–related pathways. Specifically, such small interfering RNAs and their relatives are known to direct histone modification and/or DNA methylation changes in a sequence-specific manner in yeasts, plants, nematodes, and insects (7). The current molecular models suggest that nascent gene transcripts are recognized by small RNAs through base-pairing, which brings about a local change in gene transcription through DNA methylation and histone modification.

An increasing literature implicates small RNA–mediated mechanisms in the nongenetic transmission of phenotypic traits in many organisms. In these instances, the models have small RNA amplification mechanisms that fit well with a reconstructive mode of inheritance. In some instances, specific patterns of DNA methylation and histone modification result in small RNA production that reconstructs the DNA methylation and histone state. The RNAi-related Piwi-interacting (piRNA) pathway in animals is a perfect candidate: It is conserved in mammals, including humans; is germline expressed; is known to direct histone modification and likely DNA methylation (62); and is amplified via the ping-pong pathway, among others. piRNAs themselves can be directly inherited by the offspring in invertebrates, fish, and possibly humans (63). In some paradigms of non–DNA-based inheritance in mice, sperm total RNA injection into oocytes was able to replicate paternal environmental exposure (64). However, although small RNA–based multigenerational inheritance in the nematode C. elegans has been directly demonstrated, such evidence is lacking in mammals. Other classes of RNA might be involved; for example, tRNA fragments are an abundant class of RNA in the sperm of mice and cows (65, 66). Two recent studies of heritable effects of paternal diet in mice found that the sperm small RNA fraction (67, 68) and even the tRNA fragments themselves (67) were able to phenocopy the effect of paternal diet when provided in the zygote. However, how parental diet might result in altered sperm tRNA fragment concentration or how these fragments might exert an effect in the egg remains entirely unknown.

The candidate mechanisms discussed above have the potential to share another biological function—namely, the control of the expression and mobility of transposable elements in the genome. The discovery of the first non–DNA-based alleles in plants, which are formed via a process known as paramutation (69), involved the interaction between transposable elements and genes. Many plant paramutation loci are now known to involve DNA methylation, histone modification, and repetitive genomic loci. In mice, the interindividual variation in expression at the Avy locus is modulated through variable methylation of an adjacent endogenous retroviral repetitive element (IAP), although a role for small RNAs in its control has not been shown. Finally, piRNAs and other small RNAs are important for transposable element control in many animal models, including mammals. Thus, transposable element insertions might sit at the interface between genetics and epigenetics and may generally be a source of metastable epigenetic alleles, some of which (such as Avy) might be under environmental control. These examples emphasize the importance of the interaction between genetic and epigenetic factors in apparent nongenetic inheritance and provide an important framework through which to consider candidate mechanisms.

With a number of putative candidate mechanisms in mammals having been identified, it might be tempting to measure DNA methylation states, histone modification, RNAs, etc., in population studies to identify candidate loci. However, analogous studies in plants suggest that matching phenotypes to epialleles might be challenging. Instead, a focus on a few robust, highly penetrable phenotypes might be the key to demonstrating a molecular mechanism for non–DNA sequence–based inheritance in mammals.

Are non–DNA sequence–based mechanisms adaptive, and does this matter at a population and evolutionary level?

Assuming it will be possible to clearly demonstrate new inheritance mechanisms in mammals, what might be their effects? The concept of missing heritability is sometimes evoked in this context, referring to the fact that genome-wide association studies in humans often identify genomic loci whose effect sizes cannot explain the heritable component of the phenotype in question. However, contributions of additional rarer genetic loci that require large population sizes to pass significance thresholds might provide a more parsimonious explanation. A different and useful starting point might therefore be a theoretical consideration of non–DNA sequence–based mechanisms and the constraints that they themselves might be subject to, as outlined by Rivoire and Leibler (70) (Fig. 4). Linking quantitative genetics to stochastic control theory, the authors develop a mathematical model to compare the adaptive value of classic inheritance versus an extended model that allows for information flow from the soma to the germline. These considerations suggest that, given a changing environment, the inheritance of acquired traits can be adaptive. A clear demonstration of such adaptive behavior is currently lacking for most systems. However, hypotheses such as the “thrifty phenotype” idea, whereby a compromised in utero environment programs offspring for a similarly matched environment ex utero, compellingly explains the associations between poor fetal and infant growth and increased risk of metabolic syndrome in adults born into an unmatched ex utero environment (38). But these might be considered short-term adaptations. On even longer time scales, considerations of an evolutionary role for non–DNA sequence–based mechanisms have been met with severe criticism (71), yet a potential link might be found at the interface between epigenetics and genetics. For example, the increased mutation rate at 5-methyl-cytosine in DNA compared with unmethylated cytosine (72) provides a longer-term mechanistic link between DNA sequence and other mechanisms. Another complexity to consider at the interface between genetics and epigenetics is heterozygosity. Although it is important to exclude genetic variation when establishing epigenetic mechanisms, heterozygous individuals might enhance epigenetic effects. For example, a recent study of hybrid crosses in tomato revealed an extraordinary degree of epigenetic effects (73). In this context, it is also noteworthy that environmentally associated changes in epigenetic state that depend on underlying genetic variation have recently been described in both humans and mice (74, 75).

“The idea of certain sequences that might be refractory to germline epigenetic reprogramming provides a compelling mechanism for the inheritance of modulated epigenetic states.”

Fig. 4 A generalized model of inheritance: a role in adaptation and evolution?

(A) The classic genetic model where a single attribute γ describes all heritable information that gives rise to γ′ in the offspring and that is operated on by two stochastic kernels that represent heritability (H) and reproduction (R), giving rise to ζ number of offspring. Reproduction is influenced by the environment xt at the current time. (B) A generalized model where a third stochastic kernel represents development (D) that specifies a phenotype based on an inherited genotype γ. Development, reproduction, and heritability are all under the influence of external environmental factors (xt, yt, zt). The three red arrows can be thought of as developmental plasticity (P), soma feedback to the germline (F), and Lamarckian effects (L). Certain environments might favor Lamarckian inheritance in the generalized model.

[Adapted from figure 1 of (70)]

It really is the phenomenal success of molecular genetics and a deep understanding of DNA sequence–based variation and its inheritance that make non–DNA sequence–based heritable effects seem so exotic. However, nongenetic or non–DNA sequence–based forms of inheritance have major implications for human health. It will be particularly exciting to unravel the mechanisms and their effects on humans. The tractability of nonhuman paradigms with large intergenerational or transgenerational effects will be essential for elucidating the molecular mechanisms. The consequences of delineating the interactions between genetic and epigenetic factors, even in these established paradigms, is a challenge. However, it is worth considering that even minor effects of non–DNA sequence–based inheritance might play important roles at the level of human public health.

References and Notes

  1. Acknowledgments: This work was supported by grants from Cancer Research UK (C13474/A18583 and C6946/A14492) and the Wellcome Trust (104640/Z/14/Z and 092096/Z/10/Z) to E.A.M. and from the Medical Research Council (UK) (MR/J001597/1) and the Wellcome Trust (095606/Z/11/Z) to A.C.F.-S.
View Abstract

Navigate This Article