The miR-124-Sox9 paramutation: RNA-mediated epigenetic control of embryonic and adult growth

Heritable epigenetic changes in gene expression have now been observed in a recurrent manner in the mouse. Formally comparable to the plant paramutations(Brink, 1956), they were efficiently induced in early mouse embryos by RNAs with sequence homology to the target locus, either transcript fragments or microRNAs. Carried by the sperm and ovocytes of modified parents, these RNAs act as signals of the transgenerational transfer of modified phenotypes. Paramutation of the Kit locus, first observed in the progeny of heterozygotes carrying a disrupted allele, was induced by microinjection in one-cell embryos of the cognate microRNAs miR-221 and -222(Rassoulzadegan et al., 2006). The notion of RNA-mediated epigenetic heredity was then extended to a pathological hypertrophy of the heart owing to an elevated expression of the Cdk9 transcription factor and was supported by microinjection of transcript fragments and of the cognate microRNA miR-1 into fertilized eggs(Wagner et al., 2008). In both cases, the long-term effect initiated by either the microRNA or the transcript sequences was identified as an increased transcriptional activity of the target gene, distinct from the known post-transcriptional regulations exerted by microRNAs. Taken together, these observations led to the concept of a surveillance mechanism, which, in the early embryo, detects abnormal profiles of transcripts and initiates a hereditary program enhancing expression of the normal allele at the epigenetic level.

We extended this approach to a microRNA with a distinct organ specificity, miR-124, expressed in the brain and important in the development of the central nervous system (Cao et al.,2007; Lagos-Quintana et al.,2002; Makeyev et al.,2007; Visvanathan et al.,2007). Unexpectedly, every pup born after microinjection (referred to herein as miR-124^*) showed an unusually large body size, a `giant' phenotype maintained to adulthood and subsequently inherited over several generations. The accelerated growth rate was in fact established at the most early developmental stages (morula to blastocyst). Among several transcripts upregulated in the variant embryos, our attention was drawn to Sox9 as a possible target of the paramutation. The high mobility group (HMG)-box transcription factor Sox9 is a pleiotropic actor in a number of terminal differentiation processes, including heart development, sex determination, chondrogenesis, neural crest differentiation, gliogenesis, hair follicle function, pancreas development, prostate development and retina development (Lefebvre et al.,2007; Poché et al.,2008; Thomsen et al.,2008). A crucial function of Sox9 in proliferation control in the first embryonic stem cells would be consistent with its known function in various postnatal and adult stem cells and progenitors.

RNA microinjection into the male pronucleus of fertilized B6D2 mouse eggs after spontaneous ovulation was performed by the established methods of DNA transgenesis (Hogan et al.,1994). Oligoribonucleotides and their fluorescein (FITC)-labelled derivatives were obtained from Sigma-Prolabo. The sequences used are presented in Table 1. As controls,fertilized eggs were microinjected with a 20-nt RNA oligonucleotide with an irrelevant sequence, either a fragment of the mouse Sycp1 coding sequence or other microRNAs. For studies of embryos during gestation, the control and miR-124-microinjected embryos were separately re-implanted in the left and right uterine horns of the same foster mother. Investigations were conducted in accordance with French and European rules for the care and use of laboratory animals.

Starting at day 30 postpartum, individual males were isolated with two young females in each cage. The dates of the first parturitions and, by counting backward, of the first mating were recorded. Values are the average of twelve experimental animals and twelve controls.

Analysis of early embryos was performed with the Gene Expression Cells-to-CT kit (Ambion). For late embryos and adult tissues, RNA was extracted with the Trizol Reagent (Invitrogen). 0.5 μg RNA samples were reverse-transcribed to cDNA by using random primers, hexamers and MMLV reverse transcriptase (Invitrogen). qPCR was performed using the Platinum SYBR Green qPCR SuperMix-UDG kit (Invitrogen). Sequences of oligonucleotide primers,obtained from Sigma-Prolabo, are presented Table 1.

10 μg of total RNA was loaded onto a 15% denaturing polyacrylamide gel and electrophoresed until the bromophenol blue marker reached the bottom of the gel. The separated RNA was electrotransferred to a Hybond N+ membrane(Amersham). Hybridization was carried out in the presence of a ³²P-end-labeled DNA oligonucleotide probe complementary to the mature miR-124 sequence.

ChIP was carried out using the LowCell ChIP kit (Diagenode, pAb-056-050). ChIP assays were performed with at least five embryos at day 7.5 postcoitus. After dissection, the embryos were rinsed twice in PBS. Proteins and DNA were then crosslinked with 1% formaldehyde for 8 minutes at room temperature and crosslinking was stopped with 125 mM glycine for 5 minutes. Embryos were centrifuged at 470 g in a swing-out rotor with soft deceleration settings for 10 minutes at 4°C and washed twice in 0.5 ml ice-cold PBS-butyrate solution by gentle vortexing and centrifugation as described above. Preparation of chromatin fragments (∼500 bp),immunoprecipitation and DNA recovery were performed as described in the manufacturer's procedures.

Paraffin sections (5 μm) were stained for BrdU with a monoclonal mouse antibody (Roche Molecular Biochemicals, 11170376001) diluted 1:25 in PBS, 0.1%Triton X-100 and 3% BSA; subsequent antibody detection was performed with the M.O.M. Kit (Vector Laboratories, PK-2200) and DAB (Vector Laboratories,SK-4100) as a substrate. Staining of Sox9 (Millipore, 1:100 dilution,AB5535) was performed with a biotinylated anti-rabbit antibody (Vector Laboratories), followed by incubation with peroxidase-coupled Streptavidin(Sigma) and DAB as a substrate. Nuclei were counterstained with Hematoxylin.

Embryos were fixed with 1.6% glutaraldehyde in 0.1 M cacodylate buffer then rinsed in same buffer, carefully sectioned with a razor blade, dehydrated in ethanol and treated with HMDS (hexamethyldisilazane) before air drying. Samples were coated with gold palladium and observed with a Jeol 6700F SEM.

Pregnant female mice received 50 μg/g of 5-bromo-2′-deoxyuridine(BrdU, Roche Molecular Biochemicals) by intraperitoneal injection and were sacrificed 6 hours later. Embryos were stained by immunolabelling with an anti-BrdU antibody (Roche Molecular Biochemicals) as described above.

To construct the Prm1-miR-124 transgene, pre-miR-124 was amplified by PCR using genomic DNA as a template (oligonucleotide primers revmir124 and fwdmir124). DNA oligonucleotides used to construct the Prm1-miR-124 were as follows: fwdmir124, GGACTAGTAGGCCT CTCTCTCTCCGTGTTCAC; revmir124, ATAAGAATGCGGCCGCCAGCCCCATTCTTGGCATTCA. The resulting 101-bp fragment was cloned into the pGEM-T Easy vector (Promega) and sequenced. The construct was digested with NotI and ligated into NotI digested pNASSb vector (Clonetech). The protamine-1 promoter, a gift of M. Jasin, was then cloned in front of the pre-miR-124sequence. The construct was digested with EcoRI and injected into the male pronucleus of fertilized mouse eggs(Hogan et al., 1994).

Data are expressed as mean±s.e.m. ANOVA with the Bonferroni test as a post hoc test was used versus the control. Differences between two groups were tested using the Mann-Whitney test for non-parametric samples. A P-value less than 0.05 was considered statistically significant.

In the first series of experiments, a total of 78 mice were born from fertilized eggs after microinjection in the male pronucleus of a synthetic single-stranded 22-nt oligoribonucleotide with the sequence of the mature product, identical for the three loci encoding the microRNA in the mouse genome (Mir124a-1, a-2 and a-3, Mouse Genome Informatics). Quite remarkably, all these miR-124^* mice showed a large size at birth, maintained to adulthood with body weights ∼30% greater than that of the controls (Fig. 1A,B). Such an effect had not been observed after microinjection of other microRNAs (miR-221, -222, -1, -16, -92)(Rassoulzadegan et al., 2006; Wagner et al., 2008) (our unpublished results). As shown in Fig. 1C,D, increased growth rates were steadily maintained both by males and by females during the successive phases of postnatal growth(Eisen, 1976). Most organs showed a size proportional to that of the body, with the exception of an additional increment in size of the kidney, pancreas and vertebral axis of 10-20% relative to body size (data not shown). As indicated in Fig. 1D, the males reached sexual maturity an average of 10 days before the controls.

Accelerated growth was not limited to postnatal development but was already noticeable during embryonic life. As shown in Fig. 2A-C, embryonic day 7.5(E7.5) embryos exhibited the same proportional increase in size as the newborns. In order to minimize variations in developmental timing between foster mothers, all comparative studies were performed between miR-124- and mock-injected embryos, separately transferred into the two uterine horns of the same female. Intraperitoneal injection of BrdU and immunochemical staining performed 6 hours later in miR-124^* E7.5 embryos(Fig. 2D) evidenced a larger number of labelled cells and the darker stain indicative of two successive S phases in these rapidly dividing cells(Hogan et al., 1994). Despite their larger sizes, scanning electron microscopy did not evidence a more advanced developmental stage of the E7.5 paramutant embryos(Fig. 2E).

The phenotype was already evident at the very beginning of development by the possession of a greater number of cells per blastocyst(Fig. 3A), with a frequent abnormal dispersion of the inner cell fraction, and, in some instances(∼5% of the embryos), a complete duplication of the inner cell mass(Fig. 3B). Consistent with this observation was the occurrence of a similar proportion of twin embryos linked to a common placental structure (Fig. 3C-F).

Crosses of either male or female miR-124^* with wild-type partners generated progenies with body weights significantly greater than those of the controls (Fig. 4A). In crosses with wild-type partners, transmission by both genders was efficient until the second generation, with the average weight returning to normal in the F3 progeny. The same pattern of heredity was observed in intercrosses between paramutant animals of the F0 to F2 generations (data not shown). Based on our previous observations, we considered a possible role of the gametes in the hereditary transmission of RNA molecules. Electron microscopy examination after EDTA reverse staining(Biggiogera and Fakan, 1998)evidenced increased RNA contents in miR-124^* sperm nuclei(see Fig. S1 in the supplementary material). miR-124 sequences were detected in the testis, but quantitative limitations in the recovery of sperm RNA, together with minute amounts of microRNA, made a search in spermatozoon RNA less reliable and provided variable results.

In order to evaluate the potential role of the transfer of microRNA by sperm, we developed a different strategy by first generating two transgenic families in which miR-124, expressed at the late postmeiotic stages under the control of the Prm1 (Protamine 1) promoter, accumulates in spermatozoa (Fig. 4B). All progenies of the transgenic males exhibited increased postnatal growth rates and early sexual maturity similar to those of the miR-124^*paramutants, independently of the transmission of the hemizygous transgene(Fig. 4C). In this case,however, the `giant' phenotype of the offspring was not further transmitted to the progeny by the male or female `giants'. Lack of transmission was correlated with the absence of detectable microRNA in the sperm of the non-transgenic male offspring (data not shown).

In order to identify gene(s) whose expression would be modulated in the paramutated state, we first considered the known actors of embryonic and postnatal growth, namely Gh (growth hormone), Igf1(insulin-like growth factor), Igf2 and their receptors(Fowden, 2003), but none of them was overexpressed at any stage in the paramutant embryos. We then considered the loci with sequence similarities to miR-124 previously identified as targets of the post-transcriptional regulation by the microRNA using the TargetScanHuman and miRBase databases(Karginov et al., 2007; Cheng et al., 2009). Among them, PCR assays for expression in E4.5 to E7.5 embryos led us to retain Sox9 (high-mobility-group box transcription factor), LamC1(laminin γ1 subunit), Acaa2 (acetyl-Coenzyme A acyltransferase 2) (see Fig. S2 in the supplementary material), Vamp3(Vesicle-associated membrane protein 3) and Cd164 (Cd164 antigen;data not shown). Of special interest were LamC1, required for embryonic development (Smyth et al.,1999), and Sox9, extensively studied for its crucial role in proliferation and differentiation controls in the progenitors of various organs, including those differentially affected in the paramutant pancreas,cartilage-derived skeletal structures and kidney(Lefebvre et al., 2007; Seymour et al., 2007) (A. Reginensi, M. C. Chaboissier and A. Schedl, personal communication).

Transcripts of Sox9 (Fig. 5A,B), LamC1 and Acaa2(Fig. 5B) were expressed at a low, but significant, level in the wild-type controls from as early as E2.5 up to E6.5 and showed a marked increase in miR-124^* embryos. The transcript levels were at least 5-fold higher during the pre-implantation and immediate post-implantation stages (E6.5). By contrast, accumulation of Cd164 and Vamp3 transcripts, also direct targets of the microRNA, was identical to the controls at all stages(Fig. 5B). At later stages(E15.5 to postnatal period), Sox9 RNA levels returned to levels identical in paramutants and controls. However, localized regions of increased expression of the protein were still evident in the pancreas and kidneys(Fig. 5C), the organs in which growth was more markedly increased.

Confirmatory evidence of Sox9 as a target of the paramutation was provided by microinjection assays performed with two oligoribonucleotides with sequences randomly chosen from the transcript(Table 2), as previously done in the first two instances of paramutation analyzed(Rassoulzadegan et al., 2006; Wagner et al., 2008). Both Sox9 sequences induced the oversized phenotype. Neither LamC1 nor Acaa2 transcript sequences (one randomly chosen sequence each) generated the same effect, a negative result difficult to interpret at the present stage. Further studies were then conducted on the function of Sox9 in the normal and the modified embryo.

To investigate further the mechanism of the permanent modification of Sox9 expression, we first checked whether maintenance and/or overexpression of miR-124 was involved. Not unexpectedly, given our previous observations (Rassoulzadegan et al., 2006; Wagner et al.,2008), the altered phenotype was not associated with a modified level of the microRNA. The expression of miR-124 remained identical to that of the controls, both in the embryo and in the adult(Fig. 6A,B). In the adult,expression was essentially only detected in the brain, with no significant values registered in other organs, including the kidney, in which growth was markedly affected. Quantitative PCR determination at the first stages following microinjection showed the expected high values in the zygote,quickly followed by a return to the basal level of the control(Fig. 6C). After microinjection of fluorescent FITC-tagged miR-124 oligonucleotides into the male pronucleus, the bulk of the microRNA was excluded from the nucleus within the first hour (Fig. 6D).

A heritable modification of the structure of the chromatin was evidenced at the Sox9 locus at day E6.5 after miR-124 injection by an increase in the methylated forms of histone H3 (H3K9me2 and me3), a post-translational modification previously associated with a variety of changes in genome expression, most often with silenced regions(Lachner and Jenuwein, 2002). The modification (Fig. 7)affects a putative upstream regulatory region [region `-3K' of Pan et al.(Pan et al., 2009)] in a manner characteristic of the modified embryos. The proximal promoter (+1) is loaded with the modified histones characteristic of active promoters H3K4me3 and H3Ac, but to the same extent as in the controls and modified embryos. Interestingly, the modification of the -3K region was transmitted from the founders to their progeny.

Embryo overgrowth associated with the locus-specific modulation of Sox9 expression suggests that the gene plays a central role in the control of proliferation of the embryonic stem cells in the first developmental period. Although not reported so far, such a function is not completely unexpected, as a series of terminal differentiation pathways in the adult are thought to be dependent on Sox9 activity in early progenitors.

Independent evidence of a role of Sox9 in early embryonic growth was acquired by modulating its expression at the first embryonic stages. Because of the lethality at birth of heterozygotes carrying null mutations(Bi et al., 2001), we first attempted to downregulate expression by microinjecting si-RNA molecules directed against Sox9 mRNA into zygotes. Embryos were of a small size and their development was largely abnormal, leading to extensive malformations at E10.5 and the early death expected from the E11.5 lethality of inactivating mutations. While it was clear that Sox9 is necessary during early development, a specific effect on cell proliferation and growth control could not be ascertained in this way (data not shown).

Conversely, to achieve increased levels of expression, we resorted to microinjection of a construct in which the complete Sox9 cDNA sequence is inserted downstream of the early CMV1 promoter, which drives high levels of expression in the one-cell embryo (our unpublished results). Embryos collected at E7.5 possessed, with a normal morphology, a size larger than that of the controls (Fig. 8A). Contrary to the paramutant situation, development was again eventually arrested at a later stage (E11.5), likely owing to Sox9 expression not being correctly regulated past the egg cylinder stage.

As embryogenesis proceeded, the levels of Sox9 transcripts in total RNA declined to the levels of the controls(Fig. 5B). Local sites of high expression were still noted in the organs that had a differentially higher growth rate, such as the kidney and pancreas(Fig. 5C), but this was not the case for other parts of the body, as illustrated for the lung in Fig. 5C. Increased growth rates and cell proliferation were nevertheless maintained in a coordinated manner for the whole body, raising the question of the genetic determination of the giant phenotype in late embryogenesis and postnatal growth.

One could argue that large pools of stem cells generated by accelerated growth during the first part of the pregnancy are sufficient to produce the adult `giant' phenotype. An alternative hypothesis, however, is that, as in other developmental processes (Lefebvre et al., 2007), Sox9 expression initiates a cascade of gene expression and regulatory mechanisms. Interestingly, it does not involve the genes known for their function in growth control, Gh and Igf1, whose expression was neither modified in the young nor in the adult. Two lines of evidence point to Sox8 as a possible candidate. Studies on testicular differentiation had suggested that Sox8 takes over some of the functions of its close homologue Sox9(Chaboissier et al., 2004), as Sox8-negative mutants were characterized by their small size(Sock et al., 2001). RT-qPCR assays performed on RNA from the testis, kidney and brain RNA detected a significant increase of Sox8 transcripts in the young (16.5 days post-partum) and in the adult (Fig. 8B). Although further studies are clearly required, and other genes could be involved, we hypothesize that secondary induction of Sox8 may play a role in the maintenance of accelerated growth rates up to adulthood.

The epigenetic change and the striking increase in body size established upon microinjection in the one-cell embryo of miR-124 RNA are mitotically stable during development and transmitted through meiosis and fertilization in a non-Mendelian manner. We therefore used the term`paramutation', with its initial meaning of hereditary epigenetic change, as initially observed in plants (Brink,1956). This does not, however, imply that the plant and animal phenomena are alike in every respect. Indeed, there are significant differences. The plant paramutation is essentially a gene-silencing effect(Chandler, 2007), whereas, in the mouse, it refers to transcriptional activation at the chromatin level. While non-coding RNAs were reported to control epigenetic states in plants, as in other organisms, including in the Drosophila germ line(Brennecke et al., 2008; Chambeyron et al., 2008),transgenerational determination of an epigenetic state by gametic RNA is, so far, unique to the mouse paramutation.

As in previously analyzed instances of mouse paramutation, homologous RNAs appear as the initial signal, in this case, microRNA miR-124. The subsequent maintenance of the phenotype from the blastocyst to the adult does not appear to result from a permanent activity of the microRNA. Our results,rather, indicate that initial exposure to the microRNA, and more generally to RNA with sequence homology to the transcript, results in a change in the chromatin structure of the promoter. Although the known markers of transcriptional activity, acetylation of histone H3 and trimethylation of H4(H3K4me3), were indeed detected in the promoter region (in the same amounts in both the paramutants and the controls), a discriminating and hereditary modification observed was the increase in H3K9me2 and H3K9me3 at a potential regulatory element in the upstream region of the promoter. This result was not that expected, as these variations had been associated with cases of silencing, most notably X chromosome inactivation(Lachner and Jenuwein, 2002). Several hypotheses might be considered, such as that of a repressor-RNA encoded in the upstream region. A peculiar histone modification such as H3K9 methylation might also serve different purposes in different biological contexts. A crucial piece of information that only the genetic analysis could provide is whether the upstream region is necessary for the establishment of the paramutated state.

These observations leave open the important question of the mechanism by which RNA directs the establishment of the locus of the modified chromatin structure. As recently described for X-chromosome inactivation(Zhao et al., 2008), one likely model is that the inducing RNAs target a chromatin remodelling system to the affected locus, possibly including one of the trithorax group proteins associated with transcriptional activation(Schuettengruber et al.,2007). The fact that this first encounter occurs in the secrecy of the one-cell embryo makes a search for molecular aspects technically demanding. To face this problem, we are currently working on the development of cell culture systems more amenable to molecular analysis (H. Ghanbarian,V.G. and M.R., unpublished).

A role of RNA as a transgenerational determinant was confirmed by the induction of the epigenetic state in eggs fertilized by the miR-124-loaded sperm of Prm1-transgenic males. In this case,however, the `giant' phenotype was not transmitted further by the `giant'non-transgenic progeny, whose sperm did not maintain copies of the microRNA. Not excluding other possibilities, such as differential marking of the RNA molecules or different nucleoprotein complexes, the most likely difference between the transgenic system and the paramutant mice is a smaller number of copies transferred to the oocyte via the transgenic sperm as compared with microinjection. This observation might provide a clue regarding the conditions necessary for hereditary transmission by distinguishing the requirements for epigenetic determination in the embryo and the subsequent transfer to the germ line.

Epigenetic modulation of Kit and Cdk9 expression was induced by oligoribonucleotides with sequences of their respective transcripts as efficiently as by the cognate microRNAs(Rassoulzadegan et al., 2006; Wagner et al., 2008). A search performed on this basis for genes with partial similarities to the miR-124 sequence provided a shortlist of candidates. These candidates necessarily included several of the known targets of the microRNA post-transcriptional regulation. On the basis of their early expression in development, five of them were retained for further studies, among which,three showed an elevated level of expression in the paramutant embryos. Interestingly, this was not the case of every target of the microRNA, as neither Vamp3 nor CD164 was affected. None of the three candidates at this stage, Acaa2, LamC1 and Sox9, had been reported to act in the control of embryonic growth. The mitochondrial acetyl-Coenzyme A acyltransferase 2 encoded by Acaa2 is not expected to act as a central regulator of growth, although the development of a large body implies increased metabolic activity and a corresponding contribution of mitochondria. LamC1, encoding the gamma 1 laminin subunit, and a physiological target of miR-124 during neural development(Cao et al., 2007), is necessary for basement membrane formation and development past the blastocyst stage (Smyth et al., 1999). The development of a giant embryo certainly requires basal membrane formation,but again, a central regulatory role did not appear to be likely. In fact,neither LamC1 (Table 2) nor Acaa2 oligoribonucleotides (data not shown)induced the `giant' phenotype when injected into the embryo.

The situation was different for Sox9, although its expression has not so far been recorded at the early developmental stages by analysis of expressed sequence tags (EST). Its level of expression measured by quantitative PCR in the early embryo was indeed low, but clearly significant,and even more so in the paramutated form. Epigenetic upregulation of Sox9 as a primary modification in the miR-124^*embryos would be compatible with a corresponding increase of other transcripts in a complex network of regulations, not unexpected given the multiple roles of the gene in development. A common theme of Sox9 analysis in various differentiation pathways (Lefebvre et al., 2007) is the ability to function at an early stage,possibly in the control of the proliferation and differentiation of progenitors. This is precisely the role suggested for the multipotential embryonic stem cells by the present results. It is of interest to consider that the heritability of body size in human is not fully accounted for by Mendelian determinants (Maher,2008). Inherited epigenetic determinations might offer an alternative model, especially considering the significant concentration of RNA of human sperm (Miller et al.,2005).

As it is, and as it was for mutations in classical genetics, paramutation could lead to the identification of genes involved in complex developmental phenotypes. Microinjection assays performed with other RNAs, microRNAs and their combinations might allow the discovery of new regulatory pathways. Finally, note that previously we considered(Wagner et al., 2008) that paramutation can be viewed basically as a surveillance program, recognizing an abnormal RNA profile that indicates alteration of the locus, by either illegitimate recombination or transposon insertion, and maintaining a crucial gene function by an increased expression of the correct allele.