Paramutation is a heritable epigenetic modification induced in plants by cross-talk between allelic loci. Here we report a similar modification of the mouse Kit gene in the progeny of heterozygotes with the null mutant Kittm1Alf (a lacZ insertion). In spite of a homozygous wild-type genotype, their offspring maintain, to a variable extent, the white spots characteristic of Kit mutant animals. Efficiently inherited from either male or female parents, the modified phenotype results from a decrease in Kit messenger RNA levels with the accumulation of non-polyadenylated RNA molecules of abnormal sizes. Sustained transcriptional activity at the postmeiotic stages—at which time the gene is normally silent—leads to the accumulation of RNA in spermatozoa. Microinjection into fertilized eggs either of total RNA from Kittm1Alf/+ heterozygotes or of Kit-specific microRNAs induced a heritable white tail phenotype. Our results identify an unexpected mode of epigenetic inheritance associated with the zygotic transfer of RNA molecules.
Paramutation, first observed in maize1 and subsequently in a variety of plants2, is a heritable epigenetic change in the phenotype of a ‘paramutable’ allele, initiated by interaction in heterozygotes with a ‘paramutagenic’ form of the locus. Often referred to as an exception to the law of Mendel, which states that genetic factors segregate unchanged from heterozygotes, paramutation is meiotically stable and inherited in the absence of the inducing allele. So far, the closest observations of paramutation in an animal species are changes in DNA methylation profiles directed by the allelic locus in the mouse, which we and others described as ‘transvection’ or ‘paramutation-like’ effects3,4. Here we report a modification in the phenotypic expression of the wild-type allele of the gene encoding the Kit receptor in the progeny of heterozygotes with a null insertion mutant. The ‘paramutated’ (Kit*), genotypically wild-type animals maintain the white-spotted phenotype that is characteristic of Kit mutants in the absence of the mutant allele. The efficient paternal and maternal inheritance of the paramutated state raises the question of a possible molecular support for the epigenetic information.
Non-mendelian phenotype distribution
The tm1Alf mutation (Mouse Genome Informatics accession identifier MGI:2449782; initially designated KitW-lacZ) was engineered5 by inserting a 3-kilobase (kb) lacZ-neo cassette downstream of the initiator ATG site (Fig. 1a). A unique mRNA of the same size containing the β-galactosidase coding sequence is expressed under the control of the Kit promoter and regulatory sequences. The mutation abrogates the synthesis of the Kit tyrosine kinase receptor, which has a critical role in several developmental processes including germ cell differentiation, haematopoiesis and melanogenesis. Accordingly, Kittm1Alf homozygotes die shortly after birth, and heterozygotes show a white tail tip and white feet (Fig. 1a). We initially observed an abnormal segregation of phenotypes in the progeny of crosses between two heterozygous parents. Wild-type genotypes were identified by the absence of lacZ sequences (determined by genomic polymerase chain reaction (PCR) analysis) and β-galactosidase expression (determined by in situ 5-bromo-4-chloro-3-indolyl-β-d-galactoside (X-gal) staining) (data not shown), and further confirmed by Southern blot analysis (Fig. 1c). However, it was notable that most of these genetically wild-type (Kit+/+) mice maintained the white patches characteristic of the parental heterozygotes (Fig. 1b and Table 1). The occurrence of this modified ‘paramutated’ form of the Kit+ allele (Kit* phenotype) was not restricted to the progeny of heterozygote intercrossing, but was also observed in Kittm1Alf/+ crosses with wild-type partners, independently of the gender combination (Table 1). It was not dependent on the genetic background of the mice, as the same phenotypes were found with the original 129/Sv Kittm1Alf/+ heterozygotes and after at least six generations of backcrosses of the mutation onto the C57BL/6 and B6D2 (C57BL/6 × DBA/2) genetic backgrounds (Supplementary Table 1). The paramutated phenotype was inherited with a variable phenotypic extent depending on the crosses (Supplementary Fig. 1). It was most strongly expressed in second-generation crosses between Kittm1Alf/+ heterozygotes, and still clearly recognizable in the progeny of Kit* parents in the absence of the tm1Alf allele (Kit* × Kit+/+ and Kit* × Kit* crosses; Table 1 and Supplementary Fig. 1). It would then progressively disappear in the following generations.
Another feature apparent from the data in Table 1 (see also Supplementary Fig. 1) is that in crosses between heterozygotes, the Kittm1Alf/+ genotype was generated with frequencies in the range of 50–60% instead of the mendelian two-thirds. A probable explanation is that a fraction of the paramutable alleles have been modified in Kittm1Alf/+ heterozygotes to the point of not being viable any more. The extent to which expression of the wild-type allele is altered varies between individuals, as indicated by the variable extent of the white fur patches (see Supplementary Fig. 1), and either the lack or an extensively reduced level of Kit receptor expression would not be compatible with normal development.
To probe the molecular basis of the paramutated phenotype, DNA and histone methylation were investigated in a CpG-rich region (nucleotides -31 to +219, with respect to the transcription start site) that corresponds to the minimal Kit promoter6. Cytosine methylation was examined by amplification and sequencing after bisulphite treatment7. Possible changes in chromatin structure were investigated by chromatin immunoprecipitation with antibodies directed against the lysine 4 and lysine 9 dimethylated forms of histone H3, which are associated with active and repressed chromatin, respectively8. No significant change in either cytosine or histone methylation was observed between wild-type, heterozygous and paramutated animals (data not shown). However, we cannot exclude a critical role of differential DNA or histone methylation either in a specific cell type or in one of the more distant (and not yet precisely mapped) control regions that have been inferred from the analysis of Kit mutants9.
Reduced levels of polyadenylated RNA
The white-spotted phenotype of heterozygotes with a null mutant and a wild-type allele results from a reduced level of receptor expression5. We found that this was also the case in paramutated animals. Levels of polyadenylated Kit mRNA amounting to one-half of the wild-type homozygote were determined both in Kittm1Alf/+ heterozygotes and in their paramutated (Kit*) progeny (Fig. 2a), despite the presence in the latter of two structurally normal wild-type alleles. In addition to this marked decrease in mature mRNA, Kit RNA molecules of abnormal sizes accumulated; the possible origin of these RNAs—abnormal arrest and/or initiation of primary transcription, abnormal post-transcriptional processing, and/or secondary cleavage of mature RNAs—remains to be determined. Starting in heterozygotes, distinct profiles of abnormal fragments were seen in different tissues. As shown in Fig. 2b, a prominent 0.37-kb RNA species was detected in the brain, and was identified (data not shown) as a spliced fragment of the mature Kit transcript including only exons 1 and 2. In the testis, northern analysis detected a more dispersed smear of RNA molecules of multiple sizes. As these abnormal short species were identified with a 5′ probe corresponding to the region disrupted by lacZ insertion, and, on the other hand, did not hybridize with a lacZ probe, it is clear that they were derived from the genetically wild-type allele responsible for the Kit* phenotype. They were clearly distinct from the transcript limited to the lacZ coding region expressed from the mutant allele (Fig. 2b).
RNA in Kittm1Alf/+ sperm
The next question we addressed was the nature of the signal leading to the hereditary transfer of the paramutated state, first by a comparative analysis of spermatogenesis in the heterozygote and wild-type testis. A significant difference was noted in Kit transcription levels, with higher levels (determined by transcriptional run-on assays) found in heterozygotes (Fig. 3a). Deregulation was most obvious at the late spermatogenic stages. In wild-type mouse germ cells, Kit transcription is essentially restricted to spermatogonia, with reduced levels in early meiotic cells10,11,12,13. The gene is virtually silent in the haploid phase, with the exception of a shorter RNA (tr-Kit) made from an internal promoter in the most 3′ region of the locus14. In contrast, northern blot analysis in Kittm1Alf/+ germ cells showed significant amounts of 5′ Kit RNA sequences in both round and elongated spermatids (Fig. 3b). This altered pattern of expression included an increased activity of both the upstream Kit and the internal tr-Kit promoters. The promoters of both alleles were affected, as high levels of β-galactosidase synthesis were evidenced in the haploid compartment of the tubules (Supplementary Fig. 2). These changes, clearly characteristic of the heterozygous state and at which paramutation is initiated, may be related to meiotic mispairing. Such an effect would be contrasting with the equally unexplained decrease in expression due to a perturbed synapsis that was described in a recent report15.
Most likely as a consequence of its deregulated expression, Kit RNA was detected not only in the spermatids of heterozygotes and paramutated mice, but, even more unexpectedly, in their mature epididymal sperm (Fig. 4). Both semiquantitative RT–PCR (PCR with reverse transcription) and quantitative real-time PCR on sperm extracts detected RNA sequences corresponding to the 5′ region of the Kit gene. RT–PCR performed with only 20 cycles of amplification detected Kit RNA sequences (Fig. 4a), in addition to other transcripts (including Gapdh, Prm1 and Prm2) (data not shown). These RNAs were never detected in the sperm of wild-type animals at such low cycle numbers. The presence of RNA molecules in human sperm has been reported previously (reviewed in ref. 16). However, the finding of increased amounts of Kit RNA in Kittm1Alf/+ heterozygote sperm was not expected and needed confirmation. Acridine-orange staining showed two unexpected features in the sperm of heterozygous and paramutated males (Fig. 4b). Microscopic examination revealed the accumulation of yellow-stained material in the vicinity of the sperm nuclei (Fig. 4b, right panels), presumably corresponding to RNA17. Fluorescence-activated cell sorting (FACS; Fig. 4b, left panels) analysis confirmed a significant degree of yellow staining (vertical axis). However, a more variable intensity of green staining (corresponding to DNA; horizontal axis) was indicative of a less-compact chromatin structure.
RNA-containing structures can be identified by electron microscopy using the EDTA regressive staining technique, which is based on the chelation of uranyl ions by neutral EDTA (refs 18, 19). EDTA regressive staining in an Epon resin section showed densely contrasted structures corresponding to ribonucleoprotein constituents (Fig. 4c), whereas DNA-containing structures appeared greyish or bleached (see control spermatocyte sections in Fig. 4c). Enzymatic treatment of the sections (data not shown) verified that the EDTA regressive staining was abolished after RNase treatment and remained unchanged after extensive treatment of the sections with either DNase I or pronase. Spermatozoa in sections of epididymis of Kittm1Alf/+ heterozygotes showed more heterogeneous shapes and more contrasted staining than in wild-type males (Fig. 4c). The generally light staining of the wild-type mouse sperm by the EDTA regressive staining technique, with only the rare occurrence of somewhat more-contrasted sections (Fig. 4c), may reflect a physiological low level of RNA, as reported for human sperm16. Taken together with the RT–PCR results and with the deregulated expression of Kit at the late spermatogenic stages, these observations are indicative of the presence of unusual amounts of RNA.
The presence of RNA in sperm cells led us to consider the possibility that transfer of RNA to the fertilized egg could be the signal leading to the paramutated phenotype. Although the molecular mechanisms involved would remain to be established, such a hypothesis would be consistent with recent examples of RNA molecules responsible for stable epigenetic changes (reviewed in ref. 20). A series of experiments were performed in which RNA prepared from either Kit+/+ homozygotes or Kittm1Alf/+ heterozygotes was microinjected into B6D2 one-cell embryos following standard procedures for DNA microinjection21. No toxicity was noted and, in every litter born after Kittm1Alf/+ RNA injection, close to 50% of the offspring showed the white tail tip characteristic of the heterozygote (Fig. 5). The same result was achieved after injection of either somatic (brain) RNA or RNA prepared from heterozygotic sperm. We noted, however, the occurrence of rare white-spotted mice in the control litters produced after microinjection of RNA prepared from wild-type brain, as well as after injection of irrelevant (lacZ) RNA. Nevertheless, we concluded that the white-spotted phenotype observed after microinjection of Kittm1Alf/+ RNA was a specific effect of the heterozygote RNA for two reasons. First, its frequency in control groups was significantly lower than with heterozygote RNA, with smaller areas of white fur. Second, and more significantly, the rare white tail phenotype of the controls was either very inefficiently, or not at all, transmitted to the progeny in crosses with wild-type partners, in clear contrast with the phenotype induced by Kittm1Alf/+ RNA, which was efficiently transmitted to subsequent generations (Fig. 5 and Supplementary Table 3).
Taken together with the presence of RNA molecules of abnormal sizes in preparations from heterozygotes (Fig. 2), we proposed that the paramutated state was induced by a partial degradation product of Kit RNA. We attempted to induce RNA degradation by injecting either one of the two microRNAs (miRNAs), miR-221 and miR-222, that had been identified as potentially targeting Kit mRNA in a computational survey of mammalian microRNAs22, to test whether injection of these two miRNAs could have the same effect. This experiment showed that this was indeed the case, with the white tail phenotype being induced at high frequency and being efficiently inherited (Fig. 5b; Supplementary Table 2). Exposure to microRNAs of the early embryonic genome thus seems to be sufficient to induce a permanent and heritable epigenetic change in gene expression.
Although it is tempting to think in terms of causal relationships, the mechanisms leading to the inherited modification of the paramutated phenotype, and to the similar phenotypes induced by exposure to the abnormal species of Kit RNA and to microRNAs, remain to be determined. Further characterization of the RNAs in heterozygotes, their effects when injected into zygotes, and the mechanistic aspects of the chromatin remodelling processes directed by microRNAs will hopefully lead to a more complete and better-defined picture.
The initial event inducing paramutation is not known, even in the most thoroughly investigated plant systems2. Incomplete meiotic pairing of homologous chromosomes is considered as the determining event in the epigenetic changes known as co-suppression in plants and meiotic silencing in Neurospora23. Preliminary results on two other Kit mutants would be consistent with this view. We generated a distinct insertion mutant carrying a green fluorescent protein (GFP)-neo cassette in the first intron of the Kit gene and found that paramutated progeny were generated under the same conditions and with the same frequency as in Kittm1Alf/+ crosses. In contrast, strictly mendelian phenotypic and genotypic segregations were found consistently in the offspring of the classical point-mutant KitW-v (ref. 24; data not shown).
A number of epigenetic determinations are currently under study in various systems, and a role for RNA has been suggested in several instances (reviewed in ref. 2): the induction of heritable phenotypic changes by double-stranded RNA was reported in Caenorhabditis elegans25, and an RNA cache was suggested as a carrier of genetic information in Arabidopsis26. On the other hand, one of the challenging aspects of the current results is the hereditary transmission of epigenetic states. Paternal and maternal transmission were equally efficient. Analyses of the sperm cell—a simpler structure than the oocyte—led us to conclude that it is not only a vector for the male haploid genome, but also the carrier of additional information in the form of RNA molecules. The EDTA regressive staining technique, which allows the analysis of spermatozoa at the cellular level, and the microinjection of RNA and microRNAs into fertilized eggs should be useful tools to enable a functional analysis. The mouse model might then provide a clue as to the function of the RNA molecules observed in human sperm16. The hypothesis that RNAs of paternal origin, including microRNAs, can have a role in modulating gene expression in the embryo has been recently formulated (reviewed in ref. 16), and paramutation in the Kit gene may provide a useful experimental model for further analysis.
Mice and genotyping
Kittm1Alf/+ heterozygotes were initially obtained from J. J. Panthier (Institut Pasteur, Paris). Three stocks were generated and maintained in parallel—one on the original 129/Sv genetic background, and two by crosses onto C57BL/6 and C57BL/6 × DBA/2 (B6D2) backgrounds (in both cases for at least six generations). We determined genotypes by PCR assays specific for the neo and lacZ transgenes and by Southern blot hybridization with a genomic probe. Investigations were conducted in accordance with French and European regulations for the care and use of research animals.
Southern blot analysis
Analysis was performed after cleavage with EcoRI and EcoRV enzymes as described5.
Northern analysis was performed as previously described3. Polyadenylated RNA was prepared from total RNA using the mRNA isolation kit (Roche) according to the manufacturer's instructions. The 5′ probe for the detection of Kit mRNA covered the distal part of exon 1 and exon 2, from nucleotides 69 to 374 of the complementary DNA sequence (GenBank accession number AY536430). The probe for the 3′ part of the Kit transcript and the truncated tr-Kit transcript14 extended from nucleotides 2418 to 2776 (see Supplementary Information for the nucleotide sequences). Quantification was performed by densitometric analysis of autoradiographs at various exposure times. Kit RNA levels were normalized to the level of Gapdh mRNA. Quantitative PCR assays were performed with the ABI Prism apparatus (Applied Biosystems) and the Syber Green I kit (Eurogentec). Sequences of oligonucleotide primers are provided in Supplementary Information.
Transcriptional run-on assays
Assays of transcriptional activity by radiolabelling of RNA transcripts with [α-32P]UTP in isolated nuclei27 were performed on testicular cell preparations from 10-week-old males. Five independent assays per animal were performed on two animals of each genotype. A detailed procedure is provided in Supplementary Information.
Acridine-orange staining and FACS analysis
Preparation of spermatozoa from the epididymis of wild-type, heterozygote and paramutated males were fixed in buffered 10% formaldehyde for 30 min, rinsed with phosphate-buffered saline, treated with 0.2% trypsin and 0.01% Triton X-100 for 30 min at room temperature (20 °C), washed and stained with acridine orange (Sigma) for 15 min according to published procedures17. Samples were analysed on a CAS200 cell analysis system (Becton-Dickinson).
Microinjection into fertilized eggs
RNA microinjection into B6D2 fertilized eggs (after normal ovulation) was performed by the standard techniques of DNA injection21. One to two picolitres of a 10 µg ml-1 solution of total RNA or 0.1 µg ml-1 solutions of RNA oligonucleotides in 5 mM Tris-HCl, pH 8.0, 0.1 mM EDTA were injected.
In situ determination of β-galactosidase activity
X-gal staining of β-galactosidase activity was performed as previously described28.
Mouse testes and epididymes were fixed immediately after dissection in 1.6% glutaraldehyde in 0.1 M phosphate buffer (1 h at 4 °C). They were rinsed with buffer and free aldehyde groups were blocked with 50 mM NH4Cl in PBS for 30 min at 4 °C. Specimens were dehydrated with acetone and embedded in Epon resin. RNA–protein complexes were visualized by the EDTA regressive staining technique18. Shortly after, grids were stained for 1 min with 4% aqueous solution of uranyl acetate (at 18 °C) and treated for 30 min in 0.2 M EDTA, pH 7.0. Grids were carefully rinsed with distilled water and stained for 1 min with lead citrate. Under these conditions, only RNA molecules remained stained. All grids were observed in a Philips CM12 electron microscope operating at 60 or 80 kV and equipped with a 30-µm objective aperture. All recording films were taken and treated under similar working conditions.
We are indebted to K. Thyagarajan for her participation in experimental work, J. J. Panthier for the gift of Kittm1Alf/+ mice and for helpful discussions, and K. B. Marcu and A. Schedl for help in preparing the manuscript. We thank M. Aupetit, Y. Fantei-Caujolle, J. P. Laugier, S. Pagnotta and K. Rassoulzadegan for expert technical assistance. This work was made possible by a grant to M.R. as ‘Equipe Labellisée’ of the ‘Ligue Nationale Française Contre le Cancer’.
About this article
Journal für Gynäkologische Endokrinologie/Österreich (2018)