A number of recent studies have drawn attention to the role of the paternal diet on the development of complex diseases such as obesity and type 2 diabetes. This was, in particular, suggested by the epidemiological Överkalix Cohort Study conducted on residents of an isolated community in the far northeast of Sweden 4 5 . The authors found that poor diet during the paternal grandfather’s slow growth period (i.e., before puberty, between the age of 9 and 12), increased the risk of diabetes, obesity, and cardiovascular diseases in the second-generation of offspring. Studies on the molecular mechanisms involved in such type of inheritance in humans have been hindered by factors such as social, environmental, and economical variables, and have therefore triggered the generation of rodent models.

Ng and colleagues fed male rats with a high-fat diet and looked for effects in their adult female offspring which were fed with a normal diet. These daughters exhibited early-onset of impaired insulin secretion, glucose tolerance, and normal adiposity. This was associated with altered expression levels of 642 pancreatic islet genes. Interestingly, hypomethylation at the interleukin-13 receptor-α2 (Il13ra2) locus was associated with the largest alteration in expression level 6 .

In another study, Carone and colleagues fed male mice with a low-protein diet 7 . Expression profile screening of genes in offspring of both sexes revealed elevated hepatic expression of many genes involved in lipid and cholesterol biosynthesis. Modest changes in DNA methylation were detected at many sites, including a consistent change close to the Ppara gene encoding the peroxisome proliferator-activated receptor-α, a regulator of lipid metabolism. Although molecular analyses indicated that DNA methylation might be involved in this type of regulation, the implication of other mechanisms cannot be excluded.

In brief, these studies provide evidence that diet can heritably influence gene expression patterns by modifying epigenetic profiles. Such an association between diet and epigenome might be an underlying factor in the epidemic levels of obesity and metabolic syndromes that have become major public health problems in modern day Western countries.

In recent years, studies in rodents have established a causal link between in utero exposure to endocrine disruptors and the development of kidney and prostate diseases as well as genital tract abnormalities later in adults. Furthermore, the works of Dr Skinner’s research group support that these abnormalities are, in turn, passed down to the males of subsequent generations and are associated with DNA methylation abnormalities and transcriptional alterations 8 9 . However, the data is still to be confirmed by other independent studies 10 11 .

Imprinting sensory inputs present in the environment during infancy is an adaptive behavior that is conserved in every animal species as well as in humans. Imprinting establishes a positive attachment to the native environment; it drives many species-specific behaviors at the adult stages. In the worm, odor stimuli in the young larvae environment leave life-long lasting imprints that enhance attraction of the adults for the imprinted odors 12 . Using this model, the authors were able to demonstrate that imprinting can be transmitted to the following generation. In addition, they showed that small RNAs purified from imprinted worms successfully transferred imprinting to naïve worms by the way of ingestion. Finally, they demonstrated that if the environmental change does not persist (odors removed), inheritance of adaptation is limited to one generation. In contrast, when odors were maintained in the environment during at least five generations, adaptation became fixed and innately expressed in the population. The new phenotype (imprint) is associated with RNA production and is dependent on RNAi machinery and chromatin modification effectors. An efficient system of siRNA amplification ensures the maintenance of the gene silencing even after the initial trigger is gone, allowing siRNA to be transmitted to the progeny 12 13 . Heritable RNAi defective mutants have been isolated through genetic screens that identified a number of sid (systemic interference deficient) genes, including sid-1, encoding a non-specific RNA canal, or hrde-1, encoding an Argonaute protein 12 13 14 .

In addition to diet and nutrition, environmental factors encountered in early life can also have an impact on the subsequent generations. In particular, environmental factors may produce trans-generational effects on behavior and brain functions 15 16 17 . In humans, women exposed to negative behavioral factors such as childhood neglect or abuse are more likely to have children with a higher susceptibility to psychiatric disorders later in life, even when those children are raised in normal conditions where they are not exposed to neglect or abuse 15 . This has pointed to the existence of potential molecular mechanisms that are independent of the environment during pregnancy or the maternal care and which confer disease susceptibility by transmitting older molecular alterations induced by environmental factors.

With the advent of mouse models for early trauma and stress, studies in animals have started addressing the implications of those environment-independent molecular mechanisms. One of the most relevant models is based on unpredictable maternal separation combined with unpredictable maternal stress (MSUS) in the mouse, a manipulation that mimics early chronic stress in humans. When the impact of MSUS on the offspring of the stressed animals was examined, it was found that the manipulation induced multiple behavioral alterations. It resulted in severe symptoms of depression and impulsivity, and in impaired social behaviors in the stressed animals (F1 generation) after they became adults. Furthermore, these behavioral symptoms were also observed to a similar extent in the immediate offspring (F2 generation) of the stressed animals, even though these offspring were reared normally and never exposed to any stress. The following F3 generation also exhibited similar behavioral alterations, indicating that the manipulation had had a persistent effect that could be transmitted across several generations 18 19 20 . Transmission occurred through both females and males, suggesting the existence of multiple mechanisms for epigenetic transgenerational inheritance. A greater understanding of the molecular bases of the effects of such experiences on the brain and on the germline epigenome should one day aid in the treatment of mental illnesses.

Our environment (i.e., dietary habits, chemical exposure) can positively or negatively impact our lives and the ones of our progeny. A study performed by A. Brunet’s team has demonstrated that our environment can also affect our longevity and that of our descendants. They showed that epigenetic regulators implicated in histone modifications, a type of chromatin modification, could affect the longevity of the worm. More importantly, the lifespan of the progeny was affected even when the original modification was no longer present in the descendants 21 .

Another example of transgenerational epigenetic heredity was observed at two independent loci in the mouse, the viable yellow Agouti locus (Avy) 22 and the Axin fused locus (Axin ^Fu ) 23 . The phenotype variations observed at these two loci correlated with the methylation status of transposons located upstream of the genes 24 . In Axin ^Fu , the phenotypes correlated with differential DNA methylation at a retrotransposon within Axin ^Fu and the transcription of a mutant close to the retrotransposon LTR. The parental phenotype was transmitted by both the paternal and the maternal alleles for the Avy locus, or by the maternal allele solely for the Axin ^Fu locus.

In 1956, Alexander Brink defined the paramutation as an interaction between 2 alleles of a gene that causes a heritable modification in gene expression 25 . The paramutation can be defined with 3 key characteristics. First, the new pattern of gene expression is transmitted to subsequent generations, even if the inducing allele is not present. Second, the allele which received the new epigenetic information continues to provide this new information to the other allele. Third, the nucleotidic sequence of the affected allele is not altered, indicating that an epigenetic modification is involved in this process. Although the molecular mechanisms of the paramutation remain to be clarified, an increasing number of data point to RNA molecules as crucial players. In plants, several“paramutable” loci (the expression of which can be altered by paramutations) have been analyzed 26 . In corn, the locus b1 encoding a transcription factor that drives the biosynthesis of the purple pigment anthocyanin, was the most analyzed. On the one hand, the sequences required for the paramutation, namely the repeated sequences located approximately 100 kb upstream of the promoter, were transcribed and led to the production of non-coding double-stranded RNAs important for the silencing of the targeted loci 27 . This region is differentially methylated and contains a chromatin structure that differs between the wild-type and the paramutated allele. This process is called RNA-directed DNA methylation or RdDM. Importantly, extensive screens of mutants that are deficient either in the establishment or in the maintenance of paramutations have uncovered several RNA-polymerases dependent on RNA (RdRP) that are strictly required for the transcriptional gene inactivation. While the control and the maintenance of the epigenetic states are yet to be fully understood, all of those results converge toward a leading role for RNA molecules.

The mouse paramutation was first described in the inherited epigenetic modification of expression of the Kit gene, which results in a fur-color phenotype convenient for genetic studies. In the homozygous state, null mutations of the Kit gene (a tyrosine kinase receptor) are lethal because of altered differentiation and migration of the melanoblasts. In contrast, heterozygotes show a characteristic distribution of white fur patches (“white-spotting”). Importantly, we noticed that in the progeny of heterozygote intercrosses and of crosses among wild type partners, the white patch feature was maintained in the homozygous Kit ^+/+ progeny. The two wild type alleles were structurally normal at the level of the DNA sequence, thus revealing that the molecular modifications occurred at an epigenetic level. Strikingly, similar to the plant paramutation, the mutant phenotype was efficiently inherited in the subsequent generations and biological assays support the hypothesis that RNA molecules might be the transgenerational vectors. Surprisingly, the microinjection into one-cell stage embryos of RNAs extracted from either sperm or somatic organs of heterozygote animals appeared to be efficient in induction of white fur patches phenotype in the mouse. The resulting mice exhibited the mutant kit phenotype and, strikingly, the modified phenotype was inherited throughout 3 generations thus demonstrating that RNA molecules injected into fertilized eggs were able to modify epigenetically the expression pattern of a targeted gene. These results are somehow in line with the previously discussed mechanisms of hereditary phenotypic changes inherited upon transfer of RNAs in Caenorhabditis elegans 28 . The concept of epigenetic heredity induced by RNA molecules was then extended to the microinjection of various miRs into one-cell stage embryos. For instance, microRNA miR-1 microinjection induced a heritable murine hypertrophic cardiomyopathy. In this model, the paramutation induced the transcriptional activation of Cdk9, a known main regulator of cardiac development in humans 29 . The microinjection of miR-124 was another experiment which further supported the role of RNA-mediated epigenetic inheritance in mice 30 . The miR-124 microinjected mice grew a body size 30–40 percent larger than controls and had a frequent occurrence of twin pregnancies. The phenotype was associated with the over expression of the Sox9 gene during early embryogenesis and, importantly, this epigenetic modification was transmitted over 3 generations. Given that the molecular mechanisms of RNA-mediated heredity have remained largely unknown, we hypothesized that the modifications that are crucial to the transgenerational effect may not be variations in the sequence repertoire. Our recent results took us to the still poorly explored world of RNA covalent modifications. The lead was an absolute requirement for the RNA methyltransferase Dnmt2 for the Kit* and Sox9* paramutations 31 . Unlike the other members of the Dnmt family of DNA methyltransferase, Dnmt2 is an RNA methyltransferase, tRNAs being the first recognized substrates 32 , of which methylation protects them against endonucleolytic cleavage 33 . Current speculations and our own recent results suggest a more general role in the maintenance of small noncoding RNAs.

In Drosophila, the repression of cluster of transposable elements depends on the production of piRNA (small RNAs implicated in PIWI proteins-mediated silencing of transposable elements) and induces strong trans-silencing effect (TSE). Recently, Ronsseray’s team has demonstrated that this cluster can convert other homologous transgene clusters, that are incompetent for TSE, into strong silencers, by conferring a new ability to produce piRNA. The acquired silencing capacity is very stable and can be maintained over 50 generations 34 35 .

In terms of evolution, the conservation across different species of RNA-mediated transgenerational inheritance of modified phenotypes highlights the importance of this process. The molecular mechanisms involved are not well-defined, however all the aforementioned examples (many others exist) provide evidence that RNAs are implicated in several important processes, such as transposon regulation (pi-RNA), heterochromatin formation (RdRP), developmental gene regulation, and genome stability (Table 1). Altogether, these data converge toward the same idea: RNAs can alter gene expression in such a way that the modification can be transmitted to the next generations. Whether this type of inheritance might be at work in human remains a challenging open question.