D4Z4 as a prototype of CTCF and lamins-dependent insulator in human cells

Abstract

Using cellular models that mimic the organizations of the subtelomeric 4q35 locus found in patients affected with Facio-Scapulo-Humeral Dystrophy (FSHD) and in healthy individuals, we recently investigated the biological function of the D4Z4 macrosatellite in this subtelomeric context.

We demonstrated that D4Z4 acts as a CTCF and A-type lamins dependent insulator element exhibiting both enhancer- blocking and barrier activities, and displaces a telomere towards the nuclear periphery. This peripheral positioning activity lies within a short sequence that interacts with CTCF and A-type lamins. Depletion in either of these two proteins suppresses these perinuclear activities, revealing the existence of a subtelomeric sequence that is sufficient to position an adjacent telomere to the nuclear periphery. We discuss here the biological implications of these results in the light of our current knowledge in related fields and the potential implication of other CTCF and A-type lamins insulators in the light of human pathologies.

D4Z4: A New Type of CTCF-Dependent Insulator

The human subtelomeric 4q35 locus contains an array of tandem macrosatellite repeats named D4Z4, which are linked to the autosomal dominant Facio-Scapulo-Humeral Dystrophy (FSHD) (Fig. 1A). This myopathy is caused by reduction in the number of D4Z4 below a threshold of 11 copies. Because of the complexity of the 4q35 region, the exact pathogenic mechanism remains unclear and the identity of the gene(s) implicated is still controversial but epigenetic mechanisms are likely involved.

Figure 1.

(A) Schematic representation of the 4q35 locus involved in FSHD. The D4Z4 elements are indicated (black boxes) together with different genes present in the region (grey arrows). The 10q26 telomeric locus is 98% homolog to 4q35 in its distal part but at this locus the number of D4Z4 is variable and not associated with any pathology. (B) Schematic representation of the experimental system used to investigate the positioning of D4Z4-tagged telomeres. The constructs are derived from the pCMV vector, which carries a Hygromycin resistance gene (HyTK) and an eGFP reporter. The presence of a telomere seed (arrows) allows telomeric fragmentation, a mechanism based on the non-targeted introduction of cloned telomeres into mammalian cells. The construct can induces a double strand break and the loss of the terminal chromosomal fragment followed by the elongation of a de novo telomere from the telomere seed in telomerase positive cells. Successful de novo formation of eGFP-tagged telomeres (in 90% of the hygromycin-resistant cells) or internal integration of the CMV construct were confirmed by fluorescence in situ hybridization (FISH) on metaphase spreads. The expression of eGFP is followed by flow cytometry and cells are processed according to the 3D-FISH procedure.²

In order to uncover the biological function of D4Z4 and gain further insight into the FSHD pathogenesis, we generated a range of reporter constructs that mimic the organization of the 4q35 region in FSHD patients and healthy individuals (Fig. 1B). In particular, since D4Z4 is present at the end of chromosomes, we investigated the crosstalk between this element and a telomere using a telomeric fragmentation procedure (Fig. 1B).^1,2

Thereby, we demonstrated that D4Z4 acts as a powerful insulator element exhibiting both enhancer blocking and barrier activity.¹ Interestingly, a 432 bp sequence present in the 5′ end of D4Z4 recapitulates most of these activities. An in silico search for putative protein binding sites along D4Z4 allowed us to identify a motif highly similar to known CTCF-binding sites.^3,4 Results from Electro Mobility Shift Assay (EMSA) with double-stranded labeled oligonucleotides corresponding to the CTCF binding site, chromatin immunoprecipitation experiments and knockdown of CTCF expression using RNA interference in cells transfected with constructs containing full-length D4Z4 or subfragments, demonstrate that CTCF binding is necessary for D4Z4 insulator function.¹

The 4q35 locus was consistently visualized at the periphery of the nucleus in an A-type lamins dependent manner suggesting that the particular positioning of this chromosome end could explain FSHD pathogenesis^5,6 and we tested this hypothesis in our cellular models. A-type lamins dependence of D4Z4 boundary activity was demonstrated after siRNA knockdown of LMNA expression, which led to a reliable loss of the insulator function of the whole D4Z4 or subfragments harboring anti-silencing activity. Thus, while CTCF has mostly been shown to be involved in enhancer-blocking activity,^7,8 both CTCF and A-type lamins mediate the D4Z4 insulator boundary activity against position effect.

In order to investigate the role of D4Z4 in the positioning of a telomere within the nuclear space, we used a 3D-immuno FISH technique allowing a good preservation of the nuclear ultrastructure, reconstituted the three-dimensional shape of nuclei by confocal microscopy and studied the positioning of telomeres tagged with different sequences relatively to the nuclear envelope.² We demonstrated that eGFP reporters abutting de novo formed telomeres (see Fig. 1B for methodology) reside within the innermost part of the nuclei, similarly to most natural telomeres,^2,5,6,9 while a single D4Z4 inserted at a subtelomeric position counteracts this localization and displaces a telomere towards the nuclear periphery (NP).² Other sequences tested were unable to resume such activity.

The internal positioning of most telomeric constructs is linked to the presence of telomeric repeats, since constructs devoid of telomere seed are randomly positioned. The mechanisms sustaining this particular distribution of mammalian telomeres has never been studied in details and the only data that could be relevant for such a question are the demonstrations of telomeres association to nuclear matrix components,^10,11 transient association with mobile proteins of the Inner Nuclear Membrane (INM) at the end of mitosis¹² and functional relationship between telomere length or homeostasis and lamin A variants.¹³ Recent evidence suggests that peripheral positioning and late replication might be concomitant (Arnoult et al., submitted) and that peripheral clustering of telomeres is a hallmark of senescence in human stem cells.¹⁴

Our results do not show whether A-type lamins bind directly to D4Z4 or whether CTCF and A-type lamins belong to the same complex. However, upon CTCF or LMNA transient knockdown, the peripheral positioning of the fragmented constructs is lost, showing that both proteins are essential for this activity. Interestingly, CTCF immunoaffinity chromatography followed by mass spectrometry analysis suggested that A-type lamins may interact with CTCF¹⁵ and CTCF sites are present at the border of domains associated with B-type lamins (LADs) on human chromosomes,¹⁶ suggesting that the demarcation of chromatin domains depends on the association with key components of the nuclear scaffold, a phenomenon that could be relevant also for the organization of the 4q35 region.

Classically, heterochromatin is found at the NP, which has been for long considered as a repressive compartment even if several recent reports question this simplistic view, since active regions are also observed at the nuclear rim.^17–19 Among them, some higher-order structures involving insulators have been observed in Drosophila and S. pombe. In flies, the gypsy insulator bodies are tethered to the nuclear lamina through a complex comprising lamins,²⁰ and colocalization of dCTCF with these insulator bodies suggests that dCTCF also binds to insulator bodies, in association with the nuclear matrix or the NP.^20,21 To our knowledge, D4Z4 is the first peripheral positioning element described so far in human and couples this original property with boundary activity. By analogy with gypsy, it could be hypothesized that the displacement of our D4Z4-tagged telomeres toward the nuclear periphery is related to similar mechanisms and that in eukaryotes the peripheral location of insulators correspond to the targeting to an insulation-competent subnuclear domain. A second and non-exclusive model relies on the involvement of fixed nuclear structures in barrier activity demonstrated in budding yeast, where nuclear pore components were identified in a screen for boundary activity,²² or in higher eukaryotes through the association with the nuclear matrix or the nucleolar surface.^15,23–25

Even if in human cells, structures containing CTCF and resembling insulator bodies have not been reported so far, involvement of looping in insulator function is supported by the ability of CTCF to interact with itself and form clusters.¹⁵ In our system and similar to what is described elsewhere, barrier activity could result from the physical association of the motif to a fixed nuclear substructure, such as the nuclear lamina, creating a looped structure and a physical framework for the protection against silencing.

Among the thousands of CTCF binding sites identified genome-wide, many overlap with binding sites for cohesins^26–28 and a subclass corresponds to boundaries between internal and peripheral sequences.¹⁶ Thus, considering the large number of CTCF binding sites at transitions between euchromatin- and heterochromatin,^30,31 mammalian genomes could be partitioned by CTCF in domains corresponding to chromatin loops tethered to subnuclear domains. However, we still lack strong evidence sustaining this hypothesis.

Alternatively, our CTCF and A-type lamins binding sites could cooperate with other sites, either at the 4q35 locus or within D4Z4 as suggested by the presence of another putative CTCF binding site identified in silico in the 3′ part of D4Z4 and confirmed as a CTCF binding site in vitro by EMSA (Fig. 2A). However, no CTCF enrichment was detected by ChIP and fragments containing only this site exhibited no protection against silencing. However, 3D-FISH analysis revealed that a fragment containing this putative site harbors peripheral targeting activity (Fig. 2B) suggesting that this second CTCF might not be functional, at least in the protection against position effect but may contribute to peripheral positioning and share some properties with the 5′ site.

Figure 2.

(A) In silico comparison of the D4Z4 sequence with known CTCF sites revealed the presence of two putative sites for CTCF. The first one at the 5′ end of the repeat was described in.¹ The sequence of the second site, located in the 3′ end of D4Z4 (position 2828–2839, CCGCCTCCGCGCGG) is shown in (A). To determine whether this candidate CTCF binding sequence is capable of binding to CTCF, Electro Moblity Shift assays (EMSA) were carried out as previously described.²,²⁴ Incubation of decreasing amounts of C33A nuclear extracts with labeled D4Z4 oligonucleotides (lanes 1, 2) led to the formation of a DNA-protein complex. In order to compare this site to other known CTCF sites, we used unlabeled oligonucleotides corresponding to the chicken β globin FII 5′HS4 site or the TAD1 site at the mouse TCRα-Dad1 locus²⁴ for competition assays (lanes 3 and 4). Molar excess of unlabeled FII or TAD1 displaces the binding of CTCF from the labeled D4Z4 sequence suggesting that the site at the 3′ end of D4Z4 binds CTCF in vitro. (B) Schematic representation of the D4Z4 element from position 1 to 3303 relative to the two flanking KpnI sites (K) (to scale). Fragments obtained after digestion of D4Z4 were cloned between the eGFP reporter and the telomeric seed. After transfection, cells were processed for 3D-FISH analysis in order to evaluate the positioning of the corresponding de novo formed telomere within the nuclear space. The histogram displays the mean positioning ± S.D shown by error bars of natural and fragmented telomeres within the nuclear volume, calculated from the positioning of the FISH signal from the center (0%) to the outer edge of the sphere after reduction of the outer signal (V_L = nuclear volume = 100%) until it overlaps with the FISH signal (V_l = x% of V_L). The 3D FISH analysis revealed that the 3′ end of D4Z4 might also be involved in the repositioning of a telomere toward the periphery of the nucleus since the T1XDFse construct that does not contain the proximal CTCF site also mediates peripheral positioning (B, BamHI; Bl, BlpI; E, EheI; F, FseI; K, KpnI). (C) Summary of the different regions of D4Z4 harboring anti-silencing or positioning activity.

Since the joined action of CTCF and A-type lamins has never been reported before in human cells and because of the particular features of the D4Z4 insulator especially regarding nuclear localization, we propose D4Z4 as the prototype of a new class of CTCF-dependent insulator.

Multimerization Alters D4Z4 Properties

In order to mimic the genotype of the 4q35 locus in a wide range of situations observed in FSHD patients or healthy individuals, we iteratively added groups of 4 D4Z4 repeats at the 3′ end of the eGFP gene, with or without the telomere seed and created reporter constructs with 4, 8 and 12 elements. CTCF binding and peripheral tethering are impaired upon D4Z4 multimerization and we concluded that the long D4Z4 arrays present at the 4q telomere are not responsible for its peripheral positioning but that another positioning sequence might be located elsewhere at the 4q35 locus. In agreement, dual-probes FISH analysis of the 4q35 in control (containing large arrays) and FSHD (containing shorter arrays) myoblasts using D4Z4 and D4S139 (215 kb centromeric to D4Z4) probes revealed that long D4Z4 arrays are usually less peripheral than the D4S139 marker.⁵

The mechanism that regulates CTCF binding is still unknown but a role for DNA methylation can be hypothesized. DNA methylation impairs CTCF-binding at several loci^3,4,29 and is reduced in FSHD patients.³⁰ However the impairment of CTCF binding by methylation seems unlikely in this case since the 5′ CTCF site that we defined does not contain CpG dinucleotides. Alternatively, CTCF prevents DNA methylation spreading^31,32 and the absence of CTCF on long D4Z4 arrays could be responsible for the increased methylation of the long D4Z4 array, while CTCF binding to the short arrays would prevent DNA methylation and mediate the insulation and positioning properties of the macrosatellite, leading to a gain of function of D4Z4 dependent on CTCF binding in FSHD patients.

Interestingly a recent study, describes a situation similar to ours for an array of tens of 3 kb CpG-rich macrosatellite elements (DXZ4) on the X chromosome, that binds CTCF, has enhancer blocking activity and shows differential DNA methylation.³³ The similarities between DXZ4 and D4Z4 arrays further suggest that besides unique sequences, CTCF may also contribute to genome partitioning by interacting with repetitive elements opening new grounds for the understanding of the role of epigenetic changes in the regulation of this large complement of the human genome.

D4Z4, an Evolutionary Conserved Repeat Involved in Genome Organization?

In all branches of mammals, D4Z4-related repeats can be found and may originate from the very first mammalian species over 100 millions years ago.³⁴ D4Z4 displays a high conservation in higher primates where it is organized either as arrays at orthologous subtelomeric loci or dispersed throughout the genome.^35,36 Interestingly, in higher primates, peripheral nuclear positioning is also a common property of the 4q telomere³⁷ suggesting an evolutionary conserved regulatory mechanism. It was proposed that D4Z4-like repeats arose from a retrotransposed copy of the DUXA³⁸ or DUXC³⁴ genes before the mammalian branch separated, since D4Z4 contains an ORF related to this family of double homeobox proteins. A high rate of recombination is common for subtelomeric regions.^39,40 In the human genome, D4Z4 could have evolved and gained novel functions. Interestingly, alignment of D4Z4 sequences from 4q and 10q human chromosomes, chimpanzee and orangutan reveals that some of the largest bases deletions or insertions that differ between these elements are located in the vicinity of the CTCF binding sites that we identified, but the relevance of these differences in anti-silencing, binding and peripheral tethering has to be functionally tested.

Is the Molecular Mechanism of FSHD Relevant for other Pathologies?

Our results reinforce the link between FSHD and the nuclear lamina^5,6 while also bringing CTCF into the game as a new player and we proposed a new structural model for the mechanism linking the 4q35 locus to FSHD.²

In this model, the 4q35 locus in unaffected individuals is anchored at the nuclear periphery by a proximal uncharacterized element.⁵ D4Z4 elements are methylated^30,41 but not fully packed into dense heterochromatin as suggested by the presence of active chromatin marks.^42–45 This proximal anchoring may provide a relatively high mobility of the 4q end, which has a large diffusion volume, change by the length of D4Z4 array (Fig. 3). In cells from FSHD patients, the shorter contracted D4Z4 array is not methylated and associates with CTCF and A-type lamins at the nuclear periphery. This association might limit its diffusion within the nuclear volume resulting in both cis and trans insulation of gene(s) physiologically interacting with the 4q35 terminal sequences (Fig. 3). This may lead to the dysregulation of these genes and to the FSHD phenotype. As a non-exclusive possibility, the shortened D4Z4 array could inappropriately recruit other loci to the NP, putatively through CTCF oligomerization.

Figure 3.

Model for the organization of the 4q35 locus in healthy individuals and FSHD patients. Details are given in the text.

Concerning FSHD1B (approximately 10% of patients clinically diagnosed with FSHD patients without any contraction of the D4Z4 array, OMIM 158901), D4Z4 hypomethylation could be either the cause or the consequence of CTCF-binding to some D4Z4 repeats resulting in mechanisms similar to those described above.

The role of the INM could be more complex since several lines of evidence suggest a heterogeneous structure of the nuclear envelope promoted mainly by the possible existence of peripheral micro domains with specialized activities (separated A and B-type Lamins meshworks preferentially associated with eu- and heterochromatin respectively⁴⁶ or nuclear pores). Association of CTCF and A-type lamins to the contracted D4Z4 array could relocate the 4q35 locus along the nuclear periphery to a different subdomain leading to inappropriate regulation of the FSHD gene(s). Supporting this hypothesis, similarities between transcriptomes of muscle samples from an autosomal dominant form of Emery-Dreifuss muscular dystrophy (EDMD), a muscular dystrophy linked to mutations in the gene encoding A-type Lamins (LMNA) and FSHD patients suggest common pathogenic mechanisms and some overlapping genes regulated by the interaction of the D4Z4 array with the nuclear periphery and those influenced by changes in the INM integrity. Moreover, the recent identification of NETs (nuclear envelope transmembrane proteins) specifically upregulated during muscle differentiation,⁴⁷ but of yet uncharacterized functions, is bringing up new paths to explore, considering the role of the nuclear periphery in the pathogenesis of FSHD and potentially other laminassociated myopathies.

All these results also support a genome-wide role for CTCF in shaping metazoans transcriptional maps and the partitioning of functional domains through the interaction with major components of the nuclear architecture. The main functions of CTCF have been described in several reviews^8,48 but recent examples relevant to regulation of position effect in human health will be detailed below.

Indeed, very little is known on the dysregulation of insulator elements in human diseases but emerging evidences suggest a role for CTCF in pathologies. For instance, microdeletion or microduplications of the CTCF sites at the IGF2/H19 locus are associated with some cases of non syndromic Wilm tumors.⁴⁹ Also, CTCF flanks CTG/CAG trinucleotide repeats at several disease-associated loci, such as the DM1 locus implicated in myotonic dystrophy, an autosomal dominant multisystemic disorder characterized by myotonia, muscular dystrophy, cataracts, hypogonadism, cardiac conduction anomaly and diabetes mellitus,^50,51 which is caused by a CTG repeat expansion. The physiopathology of DM1 is not fully elicited and may be linked to a complex mechanism involving chromatin changes and CTCF. Moreover, in the 3′UTR region of the DMPK gene, CTG repeats are flanked by CTCF binding-site.⁵⁰ CTCF restricts the extent of antisense transcription and constrains the spreading of heterochromatin from the triplets.^50,51 Thus, the vertebrate insulator protein might have a bivalent role around the triplets by initiating, activating and restricting bidirectional transcription, which in turn enhances and limits heterochromatin formation at the DM1 locus respectively. Similar mechanisms might also be implicated in other triplet expansion diseases.⁵²

The various examples evoked herein illustrate the importance of borders and insulators in blocking the propagation of chromatin modifications for proper gene regulation and the role of nuclear structure in this phenomenon. Thus, although never investigated in the light of CTCF function, one can speculate that hypomorphic mutations in factors associated with CTCF such as cohesins^26–28 or lamins¹⁶ might also alter insulation activity and have consequences on the regulation of chromosomal position effect (CPE) in human cells by disturbing major components of the nuclear scaffold. For instance, cohesin genes are mutated in the Cornelia de Lange syndrome, a heterogeneous developmental disorder characterized by multiple abnormalities, while mutations in the gene encoding A-type Lamins give rise to at least ten distinct, heterogeneous genetic diseases called laminopathies.⁵³ Therefore, disrupting the integrity of the complexes mediating the partitioning of functional chromatin domains may cause a wide range of defects by affecting insulators and chromatin boundaries that could lead or contribute to several multi-systemic syndromes.

Overall, our work raises new and stimulating questions on the aetiology of FSHD. The emerging new data in the field will undoubtedly lay new ground for the understanding of this complex epigenetic regulation and pathology. Furthermore, this work raises new hypotheses for the understanding of subnuclear positioning of telomeres by revealing the existence of sequences able to direct the positioning of specific sequences in collaboration with two major components of the human nucleus, CTCF and lamins. Interestingly, our findings for FSHD might also unravel new molecular mechanisms relevant to other enigmatic syndromes.