Inheritance of epigenetic variations may account for a significant part of heritability in human and in mammalian models. Heritable epigenetic variations were reported in plants under the name ‘paramutation’ more than 50 years ago. Reports by E. Whitelaw and her colleagues and by our laboratory now describe a variety of situations resulting in epigenetic inheritance in mouse systems. In the three cases that we have analysed, a transcriptional increase is initiated by RNAs related to the locus, either microRNAs or transcript fragments. RNAs carried by the spermatozoon appear as the transgenerational signals responsible for paternal transmission. Extension from mouse models to human heredity, obviously speculative at present, is encouraged by the high load of RNA in human sperm.
After more than a century of spectacular genetic advances, it may sound strange to search for mechanisms of heredity distinct from the Mendelian scheme. Still, several lines of evidence point to the suggestion that everything may not be reduced to the meiotic blending and transmission of combinations of alleles, and variation reduced to that of nucleotide sequences in DNA. A first element was the recognition that expression of either individual genes or groups of genes, even entire chromosomes, can be modified and stably maintained without changes in nucleotide sequences, either coding or regulatory. Examples and mechanisms are enumerated in the present series of Essays in Biochemistry and it is sufficient to remind the reader that these different modes of expression often correspond to distinct local chromatin structures, dependent themselves on patterns of cytosine methylation in the DNA and on the covalent modifications of histones. Structural changes corresponding to developmental and physiological regulations are often unique to a differentiation pathway and reversible. On the other hand, some are stably maintained during development, the paradigmatic case being the inactivation of one of the X chromosomes in the mammalian female, established early and maintained in a clonal manner. The ultimate stability, however, is the maintenance of epigenetic variants in successive generations, a non-Mendelian epigenetic heredity now suggested by convergent lines of evidence. One of them is the clear heritability of several biological characteristics, contrasting with the degree of variation of genomic nucleotide sequences detected by the present-day technologies, essentially GWAS (genome-wide association studies). This is, among other examples, the case of heritability of height in humans or that of familial diseases evidenced by epidemiological studies [1,2]. The ‘missing heritability’ may be explained by combinations of multiple loci with variable penetrance and each with a limited contribution. Other hypotheses will probably be needed, considering for instance some striking examples of paternal inheritance. One of them was provided by the historical records of the Överkalix community in Northern Sweden , although the generality of the results has been questioned . According to the Överkalix study, health status and food supply recorded over several generations shows the longevity of probands correlated with the food supply of the grandfathers. One hypothesis to account for the paternal transmission of an epigenetic determinant may be formulated on the basis of two recognized features of the human sperm. The first one was the presence of an important load of RNA associated with sperm chromatin , a finding that we extended to the mouse, with a variability to be discussed below. A more recent, possibly relevant, finding is that a fraction of the human sperm genome is associated with histones in a nucleosomal structure, contrary to the classical view of sperm chromatin as a homogeneous, compact and silent package of chromosomal material . The nucleosomal fraction includes several of the genes active in early development, associated with specific sets of modified histones. The possibility is open that epigenetic variations may be paternally transmitted, but it remains to be demonstrated, and this will in all likelihood not be possible in the case of human sperm. The analysis is limited to epidemiological and clinical observations, and the extensively outbred genetic structure of our species imparts a degree of uncertainty to the conclusions. Animal models are required in which inheritance can be followed in the progenies of inbred, genetically homogeneous individuals .
The first demonstration of transgenerational inheritance of an epigenetic state in the mouse was the variation of expression of alleles of the Agouti locus  dependent on transcriptional interference by active retrotransposons . More recently, we analysed in some detail several occurrences of epigenetic heredity mediated by RNA molecules. We used the term ‘paramutation’ because of similarities with a case of inherited epigenetic modification studied in plants , in spite of important differences to be discussed later.
Three independent cases of paramutation have now been analysed, modifying, respectively, expression of Kit, Cdk9 and Sox9, three genes with a major role in development [11–13]. Rather than describing one by one the ‘Kit*’, ‘Cdk9*’ and ‘Sox9*’ paramutations, we will summarize their common properties.
A first characteristic differentiates these mammalian cases from what we know in plants, essentially maize. All three of them are a transcriptional up-regulation of the affected locus, whereas the plant paramutation is essentially a silencing phenomenon , as are most of the other cases of epigenetic regulation, typically X chromosome silencing.
All three of them result in clear phenotypic changes, mimicking in several cases human pathologies. The ‘Kit* paramutants’ are the major part of the Kit+/+ progeny of Kittm1Alf/+ heterozygotes carrying a null allele created by insertion of the LacZ coding region. They maintain the phenotype of the heterozygote, namely a white-tipped tail due to abnormal migration of melanocytes. As expected from the known association of diabetes with a defective mutation of the Kit receptor gene (KitW−v) , both the Kittm1Alf/+ heterozygotes and their Kit+/+ (Kit*) paramutant progenies showed signs of diabetes (glucose intolerance), further transmitted to the progeny in crosses with wild-type partners (V. Grandjean and M. Rassoulzadegan, unpublished work). Whether or not this observation may help in understanding the known familial occurrence of the disease remains to be evaluated. Cdk9 overexpression in Cdk9* paramutants results in cardiac hypertrophy, similar to a well-known human pathology with a poor prognosis, an unknown aetiology and, interestingly, a frequent familial recurrence not clearly explained in Mendelian terms. The Sox9* paramutants show an increased body size from the first developmental period to the adult and the frequent occurrence of twin pregnancies as a result of overgrowth of the embryonic stem cells at the blastocyst stage and embryo duplication.
In all three cases, the modified phenotype was efficiently transmitted to the progeny for two to three generations of crosses with normal partners, followed by a return to the wild-type phenotype and gene expression. Transmission frequencies were non-Mendelian, always close to 100%. A key element was provided by the observation of sizable amounts of RNA in the spermatozoon head of the transmitter males. Microinjection of the sperm RNA in fertilized normal eggs resulted in approx. 50% of paramutant phenotypes.
The RNA microinjection assay allows identification of the active RNA molecules. The efficient inducers were fragments of the transcript in the form of short oligoribonucleotides and the microRNAs which target the transcript. One apparent exception was the first identification of the Kit* paramutation by the maintenance of a mutant phenotype in the genetically wild-type progeny of heterozygotes with one allele inactivated by a LacZ insert. Since, however, in this case, transcription of the recombined allele generated a truncated form of the Kit RNA, we may consider that paramutation is induced whenever abnormal forms of RNA, either truncated forms of the transcript or the microRNAs, are present at levels higher than that uniformly low in control mice. In other words, we have to postulate that a surveillance mechanism operates in the one-cell embryo, ‘the mother of the stem cells’, testing the quality of the transcript. The presence of abnormal RNAs spells danger for the gene. It may reflect an abnormal recombination event, for instance the arrival of a transposon.
Many questions are raised by these results. One asked most frequently concerns the molecular mechanisms at play. So far, we have only fragmentary answers. The proximal promoter region of the genes was not found modified one way or another in terms of cytosine methylation or histone variants. Whether changes in chromatin structure occur in unknown regulatory elements, possibly distant from the gene itself, as was found for the maize paramutation  is currently under investigation. The impact of these findings in human biology is difficult to appreciate at the present time. They clearly provide a possible functional interpretation of the presence of RNA in human sperm. It was interesting to observe that the RNA content of the sperm of paramutant males is higher than that, uniformly low, in the control mice. In situ determination by the EDTA reverse-staining technique indicated in human sperm a uniform load of RNA comparable with the highest contents observed in paramutant animals, such as the Kit heterozygotes . Purposely in these mouse experiments (see above), the controls with low sperm RNA content were the usual inbred strains homozygous for most of their genome. Hence the hypothesis that RNA accumulation in sperm could be a consequence of heterozygous states, a general feature of our species. This is only one example of the difficulties to be encountered in extending scientific conclusions to the human, a familiar if frustrating situation for the geneticist. It seems in any way most likely that sperm RNA plays significant roles in shaping embryonic development.
• Hereditary epigenetic modifications of gene expression were reported in three instances in laboratory mouse strains.
• Phenotypes were maintained and transmitted in a non-Mendelian manner, several of them reminiscent of human diseases with a familial distribution.
• The hereditary variation in gene expression, termed ‘paramutation’, can be efficiently induced by RNA homologous in sequence with the locus, either transcript fragments or microRNAs. It occurred also with high frequencies in the progeny of heterozygotes carrying a structurally modified allele generating abnormal transcripts.
• RNA molecules carried by the spermatozoon appear as the likely vectors of epigenetic heredity.
• The sperm RNA contents are low in the largely homozygous laboratory strains of mice, increased in the progeny of heterozygotes and paramutants, and high in human sperm, possibly due to the extensive heterozygous state of our genomes.
- © The Authors Journal compilation © 2010 Biochemical Society