The making of an organ

Abstract

Based initially on the observation of inheritance patterns at variance with Mendel’s first law, hereditary epigenetic variations were evidenced in the mouse. Modulating the transcription of a locus, they are induced by RNAs with sequence homology to the transcript. RNAs transferred by the gamete, including sperm, to the fertilized egg appeared to be responsible for transgenerational maintenance of the variant phenotypes. Instances of RNA-dependent variations so far analyzed in the mouse—a pathological deviation of heart development and a syndrom of gigantism initiated by hyperproliferation of embryonic stem cells—suggest a general dependence of organogenesis on epigenetic controls of gene expression.

The Epigenetic Toolbox and Inheritance

Two linked mechanisms are commonly invoked to account for variations in gene expression distinct from changes in nucleotide sequence (mutation): the methylation of cytosine in DNA and changes in chromatin structure due to covalent modifications of histones.1

The current data on DNA methylation, mostly CpG methylation in mammals, indicate that these marks in the genome are largely erased at stages from the germ cell to the early developmental stages and re-established at each generation. Changes in the DNA methylation/demethylation waves at each generation create an important possibility of variation. Experiments are still lacking to identify the memory mechanism that could maintain the methylation status of house-keeping and organ-specific genes from one generation to next. The general consensus is that CpG methylation is correlated with a reduced expression level and that genes involved in differentiation are on the whole less methylated. Thus, DNA methylation would seem to be primarily involved in the stable maintenance of the differentiation state.

Histone variants and their modifications are the other element of this important problem.2 Chromatin structure, critical to allow transcription or, alternatively, to silence specific genes, is stably maintained in differentiated somatic tissues. It is, on the other hand, with some exceptions,3 extensively remodeled in germ cells, especially the male germ cells.

Altogether, the transgenerational maintenance of specific structures and expression patterns is hard to understand, unless “memory molecules,” perhaps specialized DNA and/or RNA sequences, are dictating the distribution of the epigenetic effectors.

Signaling via RNAs: A Series of Questions

RNAs are faithfully produced from the genome. It may appear at first too simplistic to argue that RNA may be a “memory molecule.” However, it is a fascinating concept, making life more interesting if we want to think about development. With now a growing body of evidence regarding an “RNA world,” research plans can be readily devised. In order to identify an RNA molecule involved in a given regulatory process and to determine how it modifies a phenotype, the logical experimental progression is to detect in the first instance a phenotype sensitive to RNA, and then to ask: which RNA is involved. Following the classical approach of the geneticist, who made use of variant structures (in this case, mutations) to identify the critical genes, we can now use epigenetic heritable modifications (that we will describe as “paramutations”) to detect epigenetic, RNA-mediated controls. From this point onwards, a number of questions will arise. Could all classes of genes be modulated by RNA? If not, what determines the specificity of RNA sequences and which locus is affected? What is the mode of inheritance? In plants, RNA-based transmission of information has been considered because of the possibility of RNA amplification by an RNA polymerase.4 In mammals, transgenerational transfer of defined RNA species might require a covalent modification, resulting for instance in a change in stability or in intracellular traffic? Is there a possibility of RNA amplification in mammalian organisms, as the recently described RNA polymerization in a complex with telomerase?5

RNA in the Reproductive Machinery

An overall view of RNA maturation processes during germinal differentiation would help us to appreciate whether RNA molecules could be involved in transgenerational information transfer. The female gamete is more difficult to approach in this respect as its maturation essentially takes place in a small number of cells during embryonic life. On the other hand, the spermatozoon was for a long time considered as a relatively simple cell, with just one haploid genome equivalent tightly packaged in a silent form in a specialized cell capable of oriented movement and equipped to enter the peripheral structure of the oocyte. It was, however, striking to observe the presence of significant amounts of RNA in the human spermatozoon.6 At first, its role remained enigmatic. Recent results from our laboratory allow us to formulate hypotheses.

In the Developing Embryo

After fertilization, the long cell cycle of the first stages becomes shorter and shorter at the start of organogenesis. Differences are observed among species regarding the timing of activation of the zygotic transcription, in the mouse toward the end of the 2-cell stage. It means that during this very first period, the only RNAs which could play a role are the maternal and then, the paternal RNAs. So, which role(s)? We will see later that it is possible for us to interfere at these early stages by microinjecting RNA immediately post fertilization. From this earliest period, development will unfold the successive, precisely regulated stages by serial epigenetic determinations, illustrated by Waddington as a landscape with successive branching valleys extending down from a mountain’s summit.7

Despite the continuously growing amount of molecular data and the high throughput generation of gene expression profiles, we still need to make use of our imagination in order to integrate the problems and to consider a solution. Chromatin structure modifiers are major players to induce subtle diversion and differences. But what is the secret of the chromatin adaptation of a given locus? RNA guides are major candidates to bring to the locus the factors necessary for gene activation. In the fertilized egg, the pool of RNAs, either of paternal or maternal origin, would, in this view, be acting to start the organogenesis program. Later on, these initial RNAs may be lost, may be distributed by asymmetric divisions, or may still be determinant, now with the newly synthesized RNAs and proteins.

Reversibility to the Early Stem Cell Pluripotent State

Reversion to the pluripotent stem cell stage was achieved starting from a variety of differentiated cells.8 The concept of iPS cells (“induced Pluripotent Stem cells”) highlights the importance of the epigenetic processes. Reversion to the multipotent stages is achieved after either the transfer of somatic nuclei into the fertilized eggs (cloning strategy) or the transfer of the pluripotent gene expression vector in differentiated cells. Interestingly the endogenous pluripotency determinant is continuously activated in iPS cell lines in the absence of the exogenous expression vector, whose transfection had initiated the process. The epigenetic mechanisms involved have still to be discovered.

First Example of RNA Induced Hereditary Epigenetic Variation in the Mouse: A Variation in Coat Color as a Result of an Epigenetic Event in the Early Embryo

We initially started from the observation of a simple coat color pattern to show that RNA could play an unexpected role in passing characteristics on to the next generations.9 The mutant phenotype, a tail color variation, corresponding to a disrupted form of the Kit gene was hereditarily maintained even in the absence of the mutant allele. The efficient paternal heredity of the epigenetic variant phenotype led us to consider the possibility that sperm RNA would act as a transgenerational signal and we strengthened this hypothesis by showing that microinjection of this RNA into normal fertilized eggs induces the variant phenotype. This experimental approach allowed us to determine that the inducing molecules were either fragments of the transcript of the affected gene (the Kit locus) or two microRNAs known to target its transcript. We used the term “paramutation” for the mode of heritable variation thus evidenced, by analogy with a formally similar phenomenon studied in plants for a number of years.4 We do not, however, want to imply the mechanisms are identical in the two widely different types of organisms. To the contrary, it now appears that the “mouse paramutation” corresponds to an increased transcriptional activity in the locus, while the effect in plants is described as “gene silencing.”

Inducing Developmental Variation with Defined RNA Molecules: The Case of Heart Organogenesis

We further described the induction of a hereditary cardiac hypertrophy syndrome similar to human hypertrophic cardiomyopathy (HCM) by transiently changing the RNA content of the eggs via microinjection into early embryos.10 HCM is an important, sometimes hereditary, heart disease characterized by enlargement of cell size in cardiac muscle with an associated increase in RNA and protein synthesis. Familial clustering is known and clinical studies identified associated mutations, but none that would offer a general mechanism for the development of the disease, thus confirming the suspicion that epigenetic events might be involved. A variety of experimental models of cardiac hypertrophy have been developed. A major finding was the identification of key factors responsible for cardiac growth, namely a complex of proteins including Cdk9 and cyclin T1, which regulates by phosphorylation the activity of RNA polymerase II. MicroRNA miR-1 expressed in cardiac and skeletal muscle has been found to exert crucial function(s) in their development and physiology. The Cdk9 transcript, a target of miR-1, shows sequence homology with the microRNA. We examined in the HCM model the relevance of paramutation to pathophysiology and disease. Microinjection of Cdk9 transcript fragments and of the cognate microRNA miR-1 into fertilized mouse eggs efficiently induced the HCM phenotype, thereafter efficiently inherited, both paternally and maternally, in crosses with normal partners. miR-1 RNA sequences were detected in the sperm of the modified males. The HCM condition was related to an elevated expression of the Cdk9 protein. In both the Kit and the Cdk9 cases, the long-term effects initiated by either the microRNA or the transcript sequences thus correspond to an increased transcriptional activity of the target gene, distinct from the known posttranscriptional regulations exerted by the microRNAs.

Epigenetic Variation of Sox9 Expression in Early Embryonic Stem Cells: The Sox9*paramutation, Giants and Twins

We extended this approach to a microRNA with a distinct organ specificity, miR-124, expressed in the brain and with an important role in the development of the central nervous system.11 After microinjection in the zygote, an accelerated growth rate was seen, starting at the earliest developmental stages (morula to blastocyst). As a result of early growth acceleration, every baby showed at birth an unusually large body size, a “giant” phenotype maintained into adulthood and subsequently inherited over several generations. Accumulation of RNA in sperm suggested that it could act as a cross-generational signal, a point confirmed by the appearance of the giant phenotype in animals born from males expressing the microRNA during spermatogenesis. The fact that even the very early developmental stages already showed increased growth, leading frequently to duplications of the blastocyst inner cell mass accounts for the frequent occurrence of twin pregnancies. Among several transcripts upregulated in the variant embryos, our attention was drawn to Sox9 as a possible target of the paramutation. The HMG-box transcription factor Sox9 is a pleiotropic agent in a number of terminal differentiation processes, including heart development, sex determination, chondrogenesis, neural crest differentiation, gliogenesis, hair follicle function, pancreas, prostate and retina. A critical function of the gene in proliferation control in the first embryonic stem cells is in fact consistent with its known function in the various post-natal and adult stem cells and progenitors. We concluded that growth rate can be modulated at the epigenetic level.

On the Molecular Mechanisms of Paramutation: Preliminary Results and Speculations

Regarding the question frequently raised and obviously critical of the molecular mechanisms of paramutation, we still have very few hard data to offer. There are in fact two distinct questions, first that of the critical step(s) by which the presence of the RNA will trigger the epigenetic change. One obvious problem is that it occurs in the secrecy of the one-cell embryo, a rare unique cell not convenient for molecular approaches. A certainly more amenable approach would be to analyze expression in the modified tissues of the mouse. Some indications are provided by studies on other epigenetic regulatory systems. However, most of them are concerned with gene silencing rather than transcriptional activation. This is especially the case of the plant paramutation, which could appear as the closest to the “mouse paramutation,” but in fact differs on this fundamental respect.

A reasonable starting point is that a change will be eventually registered in either one of the two variables at work in epigenetic controls, cytosine methylation in DNA and covalent histone modifications in the local chromatin structures. However, the analysis of the promoter regions of the three “paramutable” genes, Kit, Cdk9 and Sox9 did not reveal a detectable variation either in DNA methylation patterns or in the histone variants most frequently associated with transcriptional activation.11,12 Current studies aim to identify regulatory elements in which such changes could be detected. Little is known, however, of the putative or actual transcription regulatory elements for either one of the three genes and, as in the case of the plant paramutation, the change in expression may depend from a still unknown element distant from the locus.

The other thought-provoking, but still rather mysterious aspect is the transfer of the RNA inducers to the spermatozoon nucleus. This may, however, be of specific interest in the biology and pathology of our species, in which epigenetic inheritance has been repeatedly conjectured.

Cases of “Missing Heritability,” the Case of the Human

Current genome-wide association studies fail to identify single loci responsible of the familial clustering of a number of diseases, thus pointing to a “missing heritability” not readily explained by Mendelian genetics.13 We offer as a partial explanation the possibility of RNA mediated hereditary variation. Previous observations6 as well as our own unpublished results point to a relatively important load of RNA in human sperm, higher than that in our control mice and comparable to that of the paramutants. Whether this may be a reflection of the widely heterozygous structure of the genomes in our fully out bred species is an attractive hypothesis that remains to be substantiated, but there is no reason to exclude a possible role of epigenetic heredity in human biology.

Concluding Remarks

We conclude that both genetic and epigenetic selection play key roles in the inheritance of variable phenotypes. The work of a number of groups points to the control of gene expression by small RNAs. We now know that such variations may be inherited. Inheritance may result from the transgenerational transfer of these molecules as well as from their synthesis during development. In addition, if small non-coding RNAs such as matured microRNAs could be transiently present in the nucleus, a control of the traffic from the cytoplasm may contribute to the adjustment of genomic expression in complex biological systems. From the evolutionary point of view, experience acquired by the organism in adjusting gene expression to optimal values in a given environment may therefore not be lost, but may be transmitted to the progeny who would again be submitted to the adaptation-selection process. A combination of genetic and epigenetic workshops would propose and natural selection select.

Acknowledgements

Work in our laboratory was funded by grants from ‘Ligue Nationale Contre le Cancer’ and ‘Agence Nationale de la Recherche (ANR-06-BLAN-0226 PARAMIR and ANR-08-GENO-011-01, EPIPATH-PARAPATH), France.