Benoit Moindrot

KLF14 acts as a master regulator of gene expression in adipose tissue and variants at the human KLF14 gene show reproducible association with type 2 diabetes and metabolic syndrome. Risk alleles are only pathological when maternally inherited, consistent with the observation that KLF14/Klf14 is a maternally expressed imprinted gene in human and mouse. However, how genomic imprinting is regulated at this important locus is currently unknown. In both species, KLF14/Klf14 is located ∼200 kb away from the paternally expressed imprinted gene MEST/Mest , regulated by a maternal gametic differentially methylated region (gDMR) at its promoter. Although the Klf14 promoter is unmethylated in most tissues, maternally inherited DNA methylation marks are paradoxically required for Klf14 expression. Here, we show that Mest and Klf14 reside within the same topologically associating domain (TAD) in embryonic stem cells (ESCs), defined by biallelic CTCF binding at the boundaries. Using allele-specific 4C-seq, we show that CTCF binding to the unmethylated Mest gDMR generates a paternal allele-specific sub-TAD encompassing Klf14 . CRISPR–Cas9 deletions of the paternal Mest promoter region in ESCs and in vivo in mutant mice result in loss of Mest expression and acquisition of biallelic expression at Klf14 . By analyzing epigenetic marks and chromatin looping in Klf14 -expressing pituitary cells, we identify a putative enhancer element shared by Mest and Klf14 , providing a mechanistic model for the regulation of Klf14 imprinting. This work defines a new role for the Mest maternally methylated gDMR, revealing that it exerts long-range effects via allele-specific modulation of TAD structures and acts as an imprinting control region in the regulation of Klf14 imprinting.

KLF14 acts as a master regulator of gene expression in adipose tissue and variants at the human KLF14 gene show strong and reproducible association with type 2 diabetes and metabolic syndrome. Risk alleles are only pathologic when maternally inherited, consistent with the observation that KLF14/Klf14 is a maternally-expressed imprinted gene in human and mouse. However, how genomic imprinting is regulated at this important locus is currently unknown. In both species, KLF14/Klf14 is located ~200 kb away from the paternally imprinted gene MEST/Mest, which is regulated by a maternal gametic differentially methylated region (gDMR) at its CpG island (CGI) promoter. Although the Klf14 CGI is kept unmethylated in most tissues, maternally-inherited DNA methylation (DNAme) marks are paradoxically required for Klf14 expression in mouse. Here, we show that Mest and Klf14 reside within the same topologically associating domain (TAD) in both mouse embryonic stem cells (ESCs) and differentiated cells, defined by sites of biallelic CTCF binding at the boundaries. Using allele-specific 4C-seq in F1 hybrid ESCs, we show that CTCF binding specifically to the unmethylated Mest gDMR generates a paternal allele-specific sub-TAD encompassing Klf14. We further show through CRISPR-Cas9-mediated mutagenesis, that deletions of the paternal Mest promoter region in ESCs as well as in vivo in mutant mice, results in both loss of Mest expression and acquisition of biallelic expression at Klf14. By analyzing epigenetic marks and chromatin looping in the Klf14-expressing pituitary cell line AtT-20, we identify a putative enhancer element shared by Mest and Klf14, providing a mechanistic model for the regulation of Klf14 imprinting. This work defines a new role for the Mest maternally methylated gDMR, revealing that it exerts long-range effects via allele-specific modulation of TAD structures and consequently act as an imprinting control region (ICR) in the regulation of Klf14 imprinting. Conservation of this TAD organisation at the human locus suggests that a similar regulatory mechanism operates at KLF14.

L'organisation de la chromatine a une importance fonctionnelle pour contrôler le programme d'expression des gènes. Par contre, les liens qui l'unissent au déroulement de la réplication de l'ADN sont beaucoup moins connus. Grâce à des approches basées sur la capture d'interactions chromosomiques et sur l'imagerie cellulaire, nous avons étudié les liens entre le repliement à grande échelle de la chromatine et le timing de réplication. Cette analyse, effectuée dans des cellules humaines lymphoblastoïdes, des cellules mononucléées du sang (PBMC) et des cellules issues d'une leucémie myéloïde à caractère érythrocytaire, a permis l'identification de domaines structuraux du noyau. Ces domaines sont relativement isolés les uns des autres et leurs frontières coïncident avec les zones d'initiation précoce. De plus, notre étude montre que ces zones d'initiation précoce interagissent préférentiellement, aussi bien entre voisins immédiats (séparation génomique de l'ordre de la mégabase) que le long du chromosome entier. Les loci répliqués tardivement interagissent eux-aussi avec leurs homologues, conduisant, dans l'espace nucléaire, à une ségrégation des loci en fonction de leur timing de réplication. Ces résultats sont soutenus par des mesures de distances sur des hybridations in-situ qui montrent que les loci répliqués en début de phase S sont plus proches qu'attendus. Nos travaux révèlent enfin que l'organisation de la chromatine est similaire dans des cellules en phase G0 (PBMC dormantes), démontrant qu'elle n'est pas spécifique des cellules en phase S. Pris ensemble, ces résultats apportent des preuves directes d'une organisation robuste de la chromatine, partagée par les cellules en cycle et dormantes, et corrélée au timing de réplication à différentes échelles.

Besides genome-wide patterns of replication timing (RT), some genes display allelic replication asynchrony in stem cells, brought about by stochastic events and genetic polymorphisms. Whether epigenetic modifications control asynchronous replication remains unclear. Here, we explore domains controlled by genomic imprinting, where parental DNA methylation imprints mediate allele-specific gene expression. Our genome-wide and locus-specific assays in mono-parental and hybrid mouse ESCs reveal pronounced RT asynchrony—which is parent-of-origin dependent and lost upon neural differentiation—at the Dlk1-Dio3 and Snrpn domains, which both comprise lncRNA polycistrons. Generating a range of mutant lines, we find that asynchronous replication at Dlk1-Dio3 is mediated by differential DNA methylation, and that the lncRNA Meg3 controls early replication across parts of the domain on the maternal chromosome. Moreover, we find no evidence that RT and organisation into TADs are linked in this domain. The combined replication timing, DNA methylation, 3D chromatin structure, and gene expression data highlight how parental methylation imprints and lncRNA expression control replication and can override RT domain organisation.

Imprinted genes provide an attractive paradigm to unravel links between transcription and genome architecture. The parental allele-specific expression of these essential genes — which are clustered in chromosomal domains — is mediated by parental methylation imprints at key regulatory DNA sequences. Recent chromatin conformation capture (3C)-based studies show differential organization of topologically associating domains between the parental chromosomes at imprinted domains, in embryonic stem and differentiated cells. At several imprinted domains, differentially methylated regions show allelic binding of the insulator protein CTCF, and linked focal retention of cohesin, at the non-methylated allele only. This generates differential patterns of chromatin looping between the parental chromosomes, already in the early embryo, and thereby facilitates the allelic gene expression. Recent research evokes also the opposite scenario, in which allelic transcription contributes to the differential genome organization, similarly as reported for imprinted X chromosome inactivation. This may occur through epigenetic effects on CTCF binding, through structural effects of RNA Polymerase II, or through imprinted long non-coding RNAs that have chromatin repressive functions. The emerging picture is that epigenetically-controlled differential genome architecture precedes and facilitates imprinted gene expression during development, and that at some domains, conversely, the mono-allelic gene expression also influences genome architecture.

The imprinted Dlk1-Dio3 domain comprises the developmental genes Dlk1 and Rtl1, which are silenced on the maternal chromosome in different cell types. On this parental chromosome, the domain's imprinting control region activates a polycistron that produces the lncRNA Meg3 and many miRNAs (Mirg) and C/D-box snoRNAs (Rian). Although Meg3 lncRNA is nuclear and associates with the maternal chromosome, it is unknown whether it controls gene repression in cis. We created mouse embryonic stem cells (mESCs) that carry an ectopic poly(A) signal, reducing RNA levels along the polycistron, and generated Rian−/− mESCs as well. Upon ESC differentiation, we found that Meg3 lncRNA (but not Rian) is required for Dlk1 repression on the maternal chromosome. Biallelic Meg3 expression acquired through CRISPR-mediated demethylation of the paternal Meg3 promoter led to biallelic Dlk1 repression, and to loss of Rtl1 expression. lncRNA expression also correlated with DNA hypomethylation and CTCF binding at the 5′-side of Meg3. Using Capture Hi-C, we found that this creates a Topologically Associating Domain (TAD) organization that brings Meg3 close to Dlk1 on the maternal chromosome. The requirement of Meg3 for gene repression and TAD structure may explain how aberrant MEG3 expression at the human DLK1-DIO3 locus associates with imprinting disorders.

Chromosome conformation capture techniques are a set of methods used to determine 3D genome organization through the capture and identification of physical contacts between pairs of genomic loci. Among them, 4C-seq (circular chromosome conformation capture coupled to high-throughput sequencing) allows for the identification and quantification of the sequences interacting with a preselected locus of interest. 4C-seq has been widely used in the literature, mainly to study chromatin loops between enhancers and promoters or between CTCF binding sites and to identify chromatin domain boundaries. As 3D-contacts may be established in an allele-specific manner, we describe an up-to-date allele-specific 4C-seq protocol, starting from the selection of allele-specific viewpoints to Illumina sequencing. This protocol has mainly been optimized for cultured mammalian cells, but can be adapted for other cell types with relatively minor changes in initial steps. Key words Circular chromosome conformation capture (4C) Nuclear organization DNA loops Topologically associating domains (TADs) High-throughput sequencing Single-nucleotide polymorphism (SNP)

Summary: Hox genes encode transcription factors that specify segmental identities along the Antero-Posterior body axis. These genes are organized in clusters, where their order corresponds to their activity along the body axis, an evolutionary conserved feature known as collinearity. In Drosophila , the BX-C cluster contains the three most posterior Hox genes, where their collinear activation incorporates progressive replacement of histone modifications, reorganization of 3D chromatin architecture and sequential activation of boundary elements and cis -regulatory regions. To dissect functional hierarchies, we compared chromatin organization in larvae and in cell lines, with a focus on the Abd-B gene. Our work establishes the importance of the Fab-7 boundary element for insulation between 3D domains marked by different histone modifications. Interestingly, we detected a non-canonical inversion of collinear chromatin dynamics at the Abd-B gene, with the active histone domain decreasing in size. This chromatin organization differentially instructed alternative Abd-B promoter use, thereby expanding the possibilities to regulate transcriptional output.

Background: X chromosome inactivation in mammals is regulated by the non-coding (nc) RNA, Xist, which represses the chromosome from which it is transcribed. High levels of the N6-methyladenosine (m6A) RNA modification occur within Xist exon I, close to the 5’ end of the transcript, and also further 3’, in Xist exon VII. The m6A modification is catalysed by the METTL3/14 complex that is directed to specific targets, including Xist, by the RNA binding protein RBM15/15B. m6A modification of Xist RNA has been reported to be important for Xist–mediated gene silencing. Methods: We use CRISPR/Cas9 mediated mutagenesis to delete sequences around the 5’ m6A region in interspecific XX mouse embryonic stem cells (mESCs). Following induction of Xist RNA expression, we assay chromosome silencing using allelic RNA-seq and Xist m6A distribution using m6A-seq. Additionally, we use Xist RNA FISH to analyse the effect of deleting the 5’ m6A region on the function of the endogenous Xist promoter. We purify epitope tagged RBM15 from mESCs, and then apply MS/MS analysis to define the RBM15 interactome. Results: We show that a deletion encompassing the entire Xist 5’ m6A region results in a modest reduction in Xist-mediated silencing, and that the 5’ m6A region overlaps essential DNA elements required for activation of the endogenous Xist promoter. Deletion of the Xist A-repeat, to which RBM15 binds, entirely abolishes deposition of m6A in the Xist 5’ m6A region without affecting the modification in exon VII. We show that in mESCs, RBM15 interacts with the m6A complex, the SETD1B histone modifying complex, and several proteins linked to RNA metabolism. Conclusions: Our findings support that RBM15 binding to the Xist A-repeat recruits the m6A complex to the 5’ Xist m6A region and that this region plays a role in Xist-mediated chromosome silencing.

Xist RNA, the master regulator of X chromosome inactivation, acts in cis to induce chromosome-wide silencing. Whilst recent studies have defined candidate silencing factors, their relative contribution to repressing different genes, and their relationship with one another is poorly understood. Here we describe a systematic analysis of Xist-mediated allelic silencing in mouse embryonic stem cell-based models. Using a machine learning approach we identify distance to the Xist locus and prior gene expression levels as key determinants of silencing efficiency. We go on to show that Spen, recruited through the Xist A-repeat, plays a central role, being critical for silencing of all except a subset of weakly expressed genes. Polycomb, recruited through the Xist B/C-repeat, also plays a key role, favouring silencing of genes with pre-existing H3K27me3 chromatin. LBR and the Rbm15/m6A-methyltransferase complex make only minor contributions to gene silencing. Together our results provide a comprehensive model for Xist-mediated chromosome silencing.