Myogenic Enhancers Regulate Expression of the Facioscapulohumeral Muscular Dystrophy-Associated DUX4 Gene

Abstract

Facioscapulohumeral muscular dystrophy (FSHD) is linked to epigenetic dysregulation of the chromosome 4q35 D4Z4 macrosatellite. However, this does not account for the tissue specificity of FSHD pathology, which requires stable expression of an alternative full-length mRNA splice form of DUX4 (DUX4-fl) from the D4Z4 array in skeletal muscle. Here, we describe the identification of two enhancers, DUX4 myogenic enhancer 1 (DME1) and DME2 which activate DUX4-fl expression in skeletal myocytes but not fibroblasts. Analysis of the chromatin revealed histone modifications and RNA polymerase II occupancy consistent with DME1 and DME2 being functional enhancers. Chromosome conformation capture analysis confirmed association of DME1 and DME2 with the DUX4 promoter in vivo. The strong interaction between DME2 and the DUX4 promoter in both FSHD and unaffected primary myocytes was greatly reduced in fibroblasts, suggesting a muscle-specific interaction. Nucleosome occupancy and methylome sequencing analysis indicated that in most FSHD myocytes, both enhancers are associated with nucleosomes but have hypomethylated DNA, consistent with a permissive transcriptional state, sporadic occupancy, and the observed DUX4 expression in rare myonuclei. Our data support a model in which these myogenic enhancers associate with the DUX4 promoter in skeletal myocytes and activate transcription when epigenetically derepressed in FSHD, resulting in the pathological misexpression of DUX4-fl.

INTRODUCTION

Facioscapulohumeral disease (FSHD) is an autosomal dominant muscular dystrophy characterized by progressive weakness and atrophy of specific muscle groups (1, 2). The genetics of FSHD are complex. The most common form of FSHD, FSHD1 (Online Mendelian Inheritance in Man [OMIM] entry 158900), is linked to contractions of a D4Z4 macrosatellite repeat array in the subtelomere of chromosome 4 at 4q35.2 (3–5). In the general healthy population, this repeat array varies between 11 and 100 D4Z4 repeats on both 4q chromosomes, whereas in FSHD1 patients the array is contracted to 1 to 10 repeats on one chromosome, with a requirement for at least one D4Z4 unit to develop disease. This telomeric region exists as two prominent alleles and a third rare allele distal to the array: 4qA, which contains a 6.2-kb β-satellite region, and 4qB are equally represented in the population while 4qC is rare (6). Only contractions on specific disease-permissive haplotypes of 4qA (the common 4qA161 variant and the rare 4qA159 and 4qA168 variants) are associated with FSHD1 (6–8). Less than 5% of cases (known as phenotypic FSHD or FSHD2; OMIM 158901) show no contraction of the D4Z4 repeats on chromosome 4 although these patients still carry at least one permissive 4qA161 allele (9–11).

Both forms of FSHD are associated with epigenetic alterations indicative of chromatin relaxation in the D4Z4 region. In FSHD1, there is a general DNA hypomethylation and loss of heterochromatic histone marks in the contracted allele, consistent with a chromatin environment that is permissive for gene expression (10, 12–17). Mutations in SMCHD1 (OMIM 614982), a chromosomal protein mediating DNA methylation, were identified as the epigenetic lesions responsible for the majority of FSHD2 cases and also as modifiers of disease severity in several severe cases of FSHD1 (18–20). The pathogenic effects of these FSHD-associated genetic and epigenetic changes in the D4Z4 array have been postulated to occur through a number of mechanisms. These include altered binding of a transcriptional repressor complex to the D4Z4 region (21–23), an acquired insulator function of contracted D4Z4 repeats (24), and loss of a regional regulatory boundary due to diminished nuclear matrix attachment (25, 26). The one feature these models have in common is derepression of genes in the 4q35 region.

The only 4q35-localized protein-coding gene consistently found to be misregulated in FSHD is DUX4 (OMIM 606009), a retrogene located within each D4Z4 repeat unit (27–30). In the DUX4 model, an FSHD-specific, alternative full-length mRNA splice form of DUX4 (DUX4-fl) is transcribed from the distal-most D4Z4 unit and stabilized by a polyadenylation signal encoded within the β-satellite region of 4qA disease-permissive alleles (27, 30). The DUX4-fl mRNA and protein were selectively detected in FSHD1 and FSHD2 but not in unaffected skeletal muscle cells and biopsy specimens (27). The DUX4 model was essentially confirmed independently using a larger collection of myogenic cells and biopsy specimens from FSHD patients; however, it was also demonstrated that certain unaffected first-degree relatives of patients can express DUX4-fl in the absence of clinical symptoms, although at significantly lower levels, indicating that modifiers of DUX4 expression and/or activity are involved in FSHD pathology (28). Although DUX4-fl is normally expressed in the testis (27), expression in somatic cells can be highly cytotoxic (31–35). In addition, DUX4-FL can activate a number of downstream target genes, many of which are also misregulated in FSHD (36). Thus, increased DUX4-FL expression in FSHD skeletal muscle is consistent with both FSHD1 and FSHD2, accounts for the permissive A-type subtelomere requirement, is detrimental to myocytes, and induces gene expression profiles found in FSHD muscle biopsy specimens. Together, these findings make increased DUX4-FL expression in skeletal muscle a prime mechanism for generating FSHD pathology.

Here, we describe the identification of two enhancers proximal to D4Z4 that upregulate DUX4-fl expression in skeletal myocytes but not in fibroblasts. Together with chromosome conformation capture (3C), chromatin immunoprecipitation (ChIP), and nucleosome occupancy and methylome sequencing (NOMe-seq) analyses, these results are consistent with a model in which myogenic enhancers associate with the DUX4 promoter in rare FSHD myocytes and, in combination with epigenetic dysregulation of the 4q35 region, drive pathological misexpression of DUX4-fl.

MATERIALS AND METHODS

Plasmids and antibodies.

The pGEM/42 plasmid containing the EcoRI fragment of a contracted D4Z4 locus from an FSHD patient was obtained from A. Belayew (29). This construct was digested with FseI and religated to generate plasmid pJ (containing only a single D4Z4 unit) (see Fig. 2). Plasmid pJΔ was constructed by digestion of pJ with NcoI and AvrII to remove ∼3 kb of proximal sequence and religation (see Fig. 2). The DUX4-interacting region 2 (DIR2) and all truncated sequences were PCR amplified from human genomic DNA and cloned into plasmid pJΔ using standard cloning methods (see Fig. 2 and 3). For 3C assays, the FRG1 genomic region and 4qA subtelomeric region were PCR amplified from human genomic DNA and cloned into plasmid pGEM-T Easy (Promega) using standard cloning methods. ChIP-grade antibodies used in this study were anti-histone H3 (ab1791; Abcam), anti-histone H3 methylated at K4 (anti-H3K4me1) (ab8895; Abcam), anti-histone H3 acetylated at K27 (anti-H3K27ac) (ab4729; Abcam), anti-histone H3 trimethylated at K27 (anti-H3K27me3) (ab6002; Abcam), anti-RNA polymerase II (Pol II) C-terminal domain (CTD) (ab5408; Abcam), and normal rabbit IgG (2729; Cell Signaling Technology). The monoclonal antibody F59, developed by F. E. Stockdale (37), was obtained from the Developmental Studies Hybridoma Bank developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biology, Iowa City, IA.

FIG 2.

Identification of DUX4 myogenic enhancer 1. (A) DUX4 expression constructs. For all constructs, D4Z4 repeats are indicated as triangles, and DUX4 exons 1 to 3 are shown. The diagnostic p13E-11 sequence is also indicated. (B) DUX4-interacting region 1 (DIR1) drives DUX4-fl expression in skeletal myocytes. Construct pJ or pJΔ was transiently transfected into C2C12 skeletal myoblasts or 3T3 fibroblasts. Cells were maintained in growth medium or differentiated for 72 h, and levels of DUX4-fl were quantitated by qRT-PCR (normalized to levels of 18S rRNA; NT, not transfected). Relative levels of myosin heavy chain (MYH1) were assessed to confirm the differentiation status of myoblasts and myocytes. Each bar represents the average of three independent qRT-PCRs performed in triplicate. *, P < 0.01; **, P < 0.001 (Student's t test). (C) Truncation series in DIR1. A DIR1 truncation series (deletions 1 to 7 versus full-length [FL]) was generated in pJ, with all positions indicated as bp upstream of the DUX4 MAL start codon (27). The matrix attachment region (MAR) (25) is indicated. (D) The ∼1.5-kb subregion 1 of DIR1 displays enhancer activity. pJ or a construct containing each DIR1 fragment (1 to 7) was transiently transfected into C2C12 skeletal myoblasts, cells were differentiated for 72 h, and levels of DUX4-fl were quantitated by qRT-PCR (normalized to levels of 18S rRNA; NT, not transfected). Each bar represents the average of three independent qRT-PCRs performed in triplicate.

FIG 3.

Identification of DUX4 myogenic enhancer 2 (DME2). (A) Truncation series in DIR2. A DIR2 truncation series (Δ1 to Δ3) was generated in pJ, with all positions indicated as bp upstream of the DUX4 MAL start codon (27). Fragment Δ1 was further truncated into subfragments A to F of ∼250 bp each (DME2 is indicated). (B and C) Subregion 1 of DIR2 (fragments B and D) displays enhancer activity. DIR2 fragments Δ1 to Δ3 or Δ1 subfragments A to F were cloned upstream of DIR1 in pJ (A). pJ or a construct containing each fragment (Δ1 to Δ3) was transiently transfected into C2C12 skeletal myoblasts or 3T3 fibroblasts. The Δ1 subfragments A to F were transfected into C2C12 myoblasts. Myoblasts were differentiated for 72 h, and levels of DUX4-fl were quantitated by qRT-PCR (normalized to levels of 18S rRNA; NT, not transfected). Each bar represents the average of three independent qRT-PCRs performed in triplicate. For fragment B or D versus all other fragments (A, C, E, and F), P < 0.01 (Student's t test).

Cell culture.

Generation of the original biopsy specimens and cultures of human skeletal muscle cells used in this study was approved by the Johns Hopkins School of Medicine Institutional Review Board. Primary adult normal human dermal fibroblasts (NHDF-Ad; Lonza) were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and FGM-2 SingleQuots (Lonza). Myogenic cultures derived from biceps muscle of an FSHD patient (17Abic) and an unaffected first-degree relative (17Ubic) were grown in Ham's F-10 medium (Mediatech, Inc.) supplemented with 20% FBS and SkGM-2 SingleQuots (Lonza) and differentiated in differentiation medium (DM) (DMEM–F-12 medium [1:1; HyClone] plus 2% horse serum [Lonza]) for 48 to 72 h prior to harvesting (28). C2C12 skeletal myoblasts and NIH 3T3 fibroblasts were grown in DMEM plus 10% FBS and antibiotics. C2C12 cells were either harvested as myoblasts or differentiated in DMEM plus 2% horse serum for 48 to 72 h prior to harvesting.

Immunocytochemistry.

Myogenic cells derived from biceps muscle of FSHD patients (07Abic, 09Abic, and 17Abic) and their unaffected first-degree relatives (07Ubic, 09Ubic, 17Ubic, and 17Vbic) were propagated on gelatin-coated four-well chamber slides (Thermo) until >90% confluent and then switched to differentiation medium for 4 to 6 days. Cells were rinsed two times with phosphate-buffered saline (PBS), fixed with 100% methanol for 10 min at room temperature (RT), washed three times with PBS, and incubated in blocking solution (4% horse serum [Gibco], 4% goat serum [Gibco], and 4% bovine serum albumin [BSA; EMD] in PBS–0.1% Triton X-100) for 60 min at RT. Cells were incubated overnight at 4°C with a 1:200 dilution in blocking solution of rabbit anti-DUX4-FL monoclonal antibody (MAb) E55 (Epitomics, Burlingame, CA). The following day cells were rinsed three times with PBS and incubated for 1 h with Alexa Fluor 594-conjugated anti-rabbit secondary antibody (Life Technologies Molecular Probes, Grand Island, NY) diluted 1:500 in blocking solution. Cells were subsequently costained using 1:10 dilution of the anti-myosin heavy chain (anti-MyHC) mouse MAb F59 with Alexa Fluor 488-conjugated secondary antibody (Molecular Probes) diluted 1:500 in blocking solution. Nuclei were stained with bisbenzimide. Expression was quantified by manually scanning the entire culture area to photograph and count all DUX4-FL-positive and/or MyHC-positive cells using a Nikon E800 system with Spot camera and software, version 4.6 (Diagnostic Instruments, Inc., Sterling Heights, MI).

Transient transfections.

For analysis of DUX4 expression constructs, cells were transfected with equal copy numbers (0.2 pmol) of DUX4 expression constructs using TransIT-LT1 transfection reagent (Mirus) according to the manufacturer's protocol. For analysis in differentiated C2C12 myocytes, 80% confluent cells were transfected with plasmid constructs in DM and allowed to differentiate for 72 h prior to harvesting. Other cells were also harvested at 72 h posttransfection.

Quantitative reverse transcriptase PCR (qRT-PCR).

Total RNAs were extracted using TRIzol (Invitrogen) and purified using an RNeasy minikit (Qiagen) after on-column DNase I digestion. Total RNA (2 μg) was used for cDNA synthesis using Superscript III Reverse Transcriptase (Invitrogen), and 200 ng of cDNA was used for quantitative PCR (qPCR) analysis as previously described (28). All data were normalized to levels of 18S rRNA (22). Oligonucleotide primer sequences are provided in Table S2 in the supplemental material.

In vitro methylation of DUX4 constructs.

Ten micrograms of pJΔ, DME1-pJΔ, pJ, or DME2-pJ was treated with 5 μl of either SssI or HpaII methyltransferase (New England BioLabs) in 100-μl reaction mixtures containing 1 μl of 32 mM S-adenosylmethionine (SAM) and the appropriate buffer for 2 h at 37°. Reaction mixtures were spiked with an additional 1 μl of SAM and 2 μl of methyltransferase and incubated for 2 h at 37°. An aliquot of each reaction mixture was subjected to digestion with the HpaII restriction enzyme (sensitive to CpG methylation), and complete methylation was verified by lack of digestion. Equal copy numbers (0.2 pmol) of in vitro-methylated or unmethylated plasmids were transiently transfected into C2C12 myoblasts, and cells were differentiated for 72 h. Levels of DUX4-fl were quantitated by qRT-PCR as described above.

ChIP.

Chromatin immunoprecipitation (ChIP) assays were performed with 17Abic and 17Ubic differentiated myocytes using the Fast ChIP method (38) with some modifications. Cells were fixed in 1% formaldehyde in DMEM for 10 min and Dounce homogenized (10 strokes) prior to sonication. Cells were sonicated for 10 rounds of 15-s pulses at 100% power output on a Branson Sonifier 450 (VWR Scientific) to shear the DNA to a ladder of ∼200 to 800 bp, and efficiency of shearing was verified by agarose gel electrophoresis. Chromatin was immunoprecipitated using 2 μg of specific antibodies or normal IgG. SYBR green quantitative PCR assays were performed for 40 cycles of 94°C for 15 s, 55°C for 30 s, and 72°C for 30 s. PCR products were analyzed on a 1.5% agarose gel to verify correct sizes of products and specificity of primer annealing. Oligonucleotide primer sequences are provided in Table S2 in the supplemental material.

NOMe-seq.

A nucleosome occupancy and methylome sequencing (NOMe-seq) assay was performed as described previously (39) using ∼6 × 10⁵ cells from differentiated 17Abic cultures. Intact nuclei were treated with 200 U of the GpC methyltransferase M.CviPI (New England BioLabs) and SAM for 8 min at 37°, followed by an additional 100 U of M.CviPI and SAM for 8 min at 37°. Following proteinase K treatment, genomic DNA was extracted and bisulfite converted using an Epitect bisulfite kit (Qiagen) according to the manufacturer's protocol. Regions of DUX4 myogenic enhancer 1 (DME1), DME2, and the MyoD core enhancer were PCR amplified using primers devoid of GpC or CpG dinucleotides and then TA cloned and sequenced to determine patterns of CpG and GpC methylation. Oligonucleotide primer sequences are provided in Table S2 in the supplemental material.

3C.

Chromosome conformation capture (3C) assays were performed as described previously (40). Briefly, experimental 3C libraries were constructed using ∼10⁷ to 10⁸ cells from cultures of differentiated 17Abic and 17Ubic myogenic cells and NHDF-Ad fibroblasts. Cells were cross-linked with formaldehyde and lysed; then chromatin was digested with BglII and ligated, which was followed by reversal of cross-links and purification of the DNA. A bacterial artificial chromosome (BAC) control library was constructed using equimolar amounts of PAC 503A8 (41), plasmid pJΔ, and plasmid pGEM-T Easy containing the FRG1 genomic region and 4qA subtelomeric region. Semiquantitative PCR was performed in triplicate in at least three independent experiments using a primer to the anchor fragment and each of the other primers across the FRG1, FRG2, and DUX4 loci (diagrammed in Fig. 6A). Band intensities from the experimental libraries were measured using Image Lab, version 4.0, software (Bio-Rad) and normalized to those from the control library to determine relative interaction frequencies. Primer sequences are provided in Table S2 in the supplemental material.

FIG 6.

Interactions of DME1 and DME2 with FSHD candidate genes. Higher-order structure of the genomic region encompassing the FRG1, FRG2, and DUX4 loci was determined using 3C assays of differentiated FSHD (17Abic) and unaffected (17Ubic) myogenic cultures. Interactions with the FRG2 and DUX4 loci were also assessed using 3C assays of primary human fibroblasts. (A) Schematic of the FRG1, FRG2, and DUX4 loci on chromosome 4, with the promoter of each gene (P), BglII restriction sites (B), and positions of 3C primers (arrows 1 to 19) shown. Genomic position on the chromosome is indicated below primers 7, 14, and 19. (B and C) The results of semiquantitative PCR on the 17Abic, 17Ubic, and primary human fibroblast (fib) experimental libraries normalized to a BAC control library. The interaction frequency is the normalized intensity of the PCR products obtained using primer 5 (B, DIR1 interactions) or primer 7 (C, DIR2 interactions) in combination with each of the other primers in the array. Data are plotted as relative interaction frequency versus genomic position on chromosome 4. Each data point represents the average and standard deviation of at least three independent experiments (each performed in triplicate), with the value using primer 6 set to 1. Positions of primers (1 to 19) are shown above the x axis, and the vertical gray bar represents the anchor fragment.

RESULTS

DUX4-FL is expressed primarily in differentiated myocytes.

DUX4-fl expression is linked to FSHD, where the primary affected tissue is skeletal muscle; therefore, we investigated the myogenic regulation of DUX4 expression. Our previous work showed that DUX4-FL is expressed primarily, although not exclusively, in differentiated FSHD myogenic cultures (28). Intriguingly, both mononucleated and multinucleated cells express DUX4-FL (27, 28; also data not shown); however, very little is known about the regulation of DUX4 expression, and the identities of these DUX4-FL-positive cells were not examined. Immunostaining analysis of human myogenic cultures revealed that DUX4-FL protein was found almost exclusively in differentiated myocytes identified by expression of sarcomeric myosin heavy chain (MyHC) (see Fig. S1 in the supplemental material). We examined multiple proliferating and differentiating cultures of biceps-derived FSHD (donors 07Abic, 09Abic, and 17Abic) and unaffected (donors 07Ubic, 09Ubic, 17Ubic, and 17Vbic) myogenic cells (see Table S1 in the supplemental material) (28, 42) by double immunostaining using MAb E55, which is specific for the C-terminal extension of DUX4-FL that distinguishes it from the shorter DUX4-S (27, 43), in combination with MAb F59, which reacts with all human MyHC isoforms (37, 44). Consistent with previous work, we found DUX4-FL-positive nuclei in cells from both FSHD and unaffected donors (28, 45). In total our analysis identified 826 cells that contained DUX4-FL-positive nuclei, of which 824 (99.76%) also expressed MyHC (see Fig. S1 in the supplemental material). Both mononucleated myocytes and multinucleated myotubes were included among the cells that coexpressed DUX4-FL and MyHC. In differentiating cultures, DUX4-FL expression reached its highest level after ∼4 to 6 days in differentiation medium (data not shown). In multinucleated myotubes, we commonly observed a gradient of DUX4-FL staining intensity among neighboring nuclei (see Fig. S1), as seen previously (27, 28). Because DUX4-FL expression was overwhelmingly restricted to MyHC-positive, differentiated myocytes, we report DUX4-FL expression frequency as the number of DUX4-FL-positive nuclei per 1,000 nuclei in MyHC-positive cells. In one group of assays of cultures at 4 days of differentiation, the number of DUX4-FL-positive nuclei per 1,000 nuclei in MyHC-positive cells was 0.2 ± 0.1 for 07Abic, 2.1 ± 0.8 for 09Abic, and 4.8 ± 1.0 for 17Abic cultures (average ± standard error [SE]; n = 4), results which are consistent with previous work (28). Thus, although low levels of DUX4-fl mRNA can be detected in some cultures of proliferating myocytes (27, 28), immunostaining indicates that DUX4-FL protein expression is virtually undetectable in proliferating, MyHC-negative myoblasts. We conclude that DUX4-FL expression is induced during myogenic differentiation.

D4Z4 proximal regions enriched for enhancer marks interact with DUX4 in FSHD myocytes.

Since DUX4-fl expression is induced during muscle differentiation, we sought to identify myogenic cis regions regulating DUX4 by chromosome conformation capture (3C) analysis (40). 3C has become an established technique that is widely used to identify or confirm regions of long-distance genomic interaction (40). For these and the following experiments, we used myogenic cells from an FSHD1 subject (17Abic) and an unaffected first-degree relative (17Ubic) (28, 42, 46). Differentiated 17Abic myogenic cultures have been shown to consistently express DUX4-fl, which is undetectable in differentiated 17Ubic myocytes (28). Interactions between the DUX4 promoter and ∼60 kb of upstream sequence in differentiated 17Abic myocytes were scanned using a series of PCR primers flanking BglII restriction sites with the DUX4 promoter-containing BglII fragment as the anchor fragment and normalized to a BAC control library (Fig. 1). Because interaction frequencies tend to diminish with increasing distance from the anchor fragment, true interactions are identified as reproducible, local peaks across a genomic region. We observed a peak of interaction between DUX4 and FRG2 (OMIM 609032) (Fig. 1, primers 11 and 12), which has been reported previously and attributed to an association between the FRG2 promoter and an enhancer residing in the D4Z4 repeats (47). Additionally, our analysis disclosed a novel peak of DUX4 interaction in the region covered by primer 7. Data released from the ENCODE (Encyclopedia of DNA Elements) consortium (http://genome.ucsc.edu/cgi-bin/hgGateway) indicate that this region is enriched for both the enhancer mark H3K4me1 and the active enhancer mark H3K27ac, and it also displayed a local peak of DNase I hypersensitivity in several cell types, including skeletal myocytes (Fig. 1). Since interaction frequency increases with diminishing genomic distance between two sequences, 3C is not a reliable technique for uncovering specific interactions between regions separated by only several kb. However, although we only observed a plateau in the region covered by primer 5, a small local peak was consistently observed when the target and anchor fragments were reversed (see Fig. 6). This region also displayed a small local enrichment of H3K4me1, H3K27ac, and DNase I-hypersensitive sites (Fig. 1), suggesting that it might contain an active enhancer. There appeared to be a strong interaction peak in the fragment covered by primer 2; however, this region within the DUX4 gene body directly flanks the region covered by the anchor primer and lacked enrichment of histone marks, although there is a peak of H3K4me1 and H3K27 acetylation ∼2 kb distal to it (Fig. 1). Thus, this region was not further investigated, and we focused our immediate attention on the genomic regions covered by primers 5 and 7, which are referred to as DUX4-interacting region 1 (DIR1) and DUX4-interacting region 2 (DIR2), respectively.

FIG 1.

Two D4Z4 proximal regions interact with DUX4 in FSHD myocytes. Higher-order structure of the genomic region encompassing ∼60 kb upstream of the DUX4 locus was determined using 3C assays on differentiated 17Abic (FSHD) myogenic cultures. The results of semiquantitative PCR on the 17Abic experimental library normalized to a BAC control library are shown. The interaction frequency is the normalized intensity of the PCR products obtained using primer 3 in combination with each of the other primers in the array. Data are plotted as relative interaction frequency versus genomic position on chromosome 4. Positions of BglII sites and primers (2 to 14) are indicated above the graph, and the vertical bar at primer 3 represents the anchor fragment. Peaks of relative interaction frequency with primers 11 and 12 and primer 7 indicate interaction of DUX4 with FRG2 and an uncharacterized region (DIR2), respectively. 3C data are aligned with 70-kb UCSC (University of California, Santa Cruz) Genome Browser human genome build 19 (hg19) views (chromosome 4 [chr4] positions 190930000 to 191000000) for H3K4me1, H3K27ac, and DNase I hypersensitivity (HS) that encompass the DUX4 and FRG2 loci in differentiated skeletal myocytes (http://genome.ucsc.edu/cgi-bin/hgGateway). Regions of novel interaction and/or local enrichment of histone marks and open chromatin are boxed and labeled DIR1 and DIR2. Positions of ChIP amplicons (Fig. 4) are shown as black bars above the genome browser views (in order from 5′ to 3′: upstream flanking region, DME2, DME1, the DUX4 promoter/transcription start site, DUX4 intron 1, and the pLAM region). Positions of DMEs (Fig. 2 and 3) are indicated (DME2 and DME1, respectively).

Identification of DUX4 myogenic enhancer 1 (DME1) and DME2.

To analyze DIR1 for myogenic enhancer activity, we first used a modified version of the pGEM/42 plasmid (29), called pJ (Fig. 2A). The pJ plasmid contains a fragment from the contracted chromosome 4qA of an FSHD patient that consists of one D4Z4 repeat unit encoding the DUX4 proximal promoter and exons 1 and 2, exon 3 encoding the DUX4 mRNA polyadenylation site, and ∼5,700 bp of array proximal sequence including the distal half of DIR1 (Fig. 2A). This construct is capable of expressing the pathogenic form of DUX4 (DUX4-fl) in transiently transfected C2C12 cells. MatInspector analysis of the portion of DIR1 sequence contained within pJ disclosed a number of transcription factor binding motifs commonly found in muscle-specific genes (see Fig. S2A in the supplemental material). Therefore, we shortened the proximal sequence of pJ by ∼3 kb to effectively remove the DIR1 region and generated pJΔ (Fig. 2A); then we compared DUX4-fl expression from constructs containing (pJ) and lacking (pJΔ) the DIR1 sequence in C2C12 skeletal myocytes. Since DUX4 is restricted to the primate and Afrotheria lineages, murine C2C12 cells represent a convenient muscle cell culture system in which to assay exogenous DUX4 expression without concern for endogenous gene expression. The pJ and pJΔ plasmids were transiently transfected into C2C12 skeletal myoblasts, the cells were allowed to differentiate for 72 h, and levels of DUX4-fl were quantitated by qRT-PCR using isoform-specific primers. Removal of the DIR1-containing proximal sequence in pJ (pJΔ) resulted in a 3.3-fold decrease in expression of DUX4-fl in differentiated skeletal myocytes, whereas loss of this sequence increased levels of DUX4-fl in proliferating myoblasts and in 3T3 fibroblasts (Fig. 2B). Importantly, although the DUX4 promoter in pJΔ was active in both proliferating myoblasts and fibroblasts, the DIR1-containing proximal sequence in pJ reduced this activity by ∼8-fold and ∼1.6-fold, respectively. These results indicated that DIR1 is able to induce DUX4-fl expression in differentiated myocytes while repressing DUX4-fl in proliferating myoblasts and fibroblasts.

To define the transcriptional enhancer within DIR1, we generated a truncation series in pJ (Fig. 2C) and assessed levels of DUX4-fl in C2C12 myocytes transfected with each construct. In differentiated skeletal myocytes, the ∼1.5-kb fragment 1 displayed the highest activity, driving ∼3.5-fold higher levels of DUX4-fl than the full-length sequence (Fig. 2D). Thus, fragment 1 contained the active enhancer region in concert with the DUX4 promoter, and we refer to this sequence as DUX4 myogenic enhancer 1 (DME1).

To identify a potential enhancer within DIR2, we cloned a series of ∼1-kb fragments spanning the upstream portion of DIR2 that was enriched for enhancer marks (Fig. 1) and then individually inserted each fragment upstream of DIR1 in pJ (DIR2-pJ series) (Fig. 3A). We then assessed levels of DUX4-fl in C2C12 myocytes and 3T3 fibroblasts transfected with these constructs. While all DIR2 fragments were active in differentiated skeletal myocytes, only fragment 1 displayed both high activity in myocytes (2.8-fold higher activity than pJ alone) and a corresponding lack of activity in 3T3 fibroblasts (no increase over pJ alone) (Fig. 3B). Upon further truncation of fragment 1 (Fig. 3A, subfragments A to F), the active portion was found to be localized to an ∼650-bp region containing subfragments B and D (Fig. 3C). We refer to this sequence as DUX4 myogenic enhancer 2 (DME2).

DME1 and DME2 are enriched for enhancer marks in FSHD and control myocytes.

ENCODE data indicated that the DUX4-interacting regions DIR1 and DIR2 are enriched for enhancer marks in skeletal myocytes (Fig. 1). Having identified the functional enhancers within these regions, we noticed that while DME1 overlaps a peak of H3K4me1 enrichment, DME2 lies just proximal to the strong peak of H3K4me1 and H3K27ac enrichment (Fig. 1). Thus, we performed chromatin immunoprecipitation (ChIP) assays to determine whether enhancer marks are enriched at DME1 and DME2 in differentiated cultures of 17Abic (FSHD) and 17Ubic (control) myocytes. As predicted by alignment with the ENCODE data, the enhancer mark H3K4me1 is slightly enriched at both DME1 and DME2 compared to the levels in the DUX4 promoter and flanking regions (FR and pLAM) in both FSHD and control myocytes (Fig. 4A). Interestingly, in both cell types, DME1 displays attributes of a bivalent enhancer, with enrichment of both the active H3K27 acetylation mark and the repressive H3K27me3 mark (Fig. 4B and C). In contrast, neither of these marks is enriched at DME2 (Fig. 4B and C). RNA polymerase II has been shown to be present on a subset of enhancers, where it facilitates the transcription of noncoding RNAs (ncRNAs) (48). We found that Pol II occupancy at DME1 is similar to the level at the DUX4 promoter but ∼2-fold higher at DME2 than at the DUX4 promoter in both FSHD and control myocytes (Fig. 4D). Taken together, our 3C, qRT-PCR, and ChIP results with the available ENCODE data indicated that DME1 and DME2 are functional myogenic enhancers of DUX4.

FIG 4.

DME1 and DME2 display enhancer marks in FSHD and normal skeletal myocytes. ChIP assays were performed using differentiated FSHD (A) and control (U) myogenic cultures and antibodies specific for H3K4me1, H3K27ac, H3K27me3, or RNA Pol II. (A to C) Immunoprecipitated chromatin was analyzed by qPCR using primers specific for DME1, DME2, the DUX4 promoter (Pro), the pLAM region (LAM), or an upstream flanking region (FR). Data are presented as fold enrichment of the target region by each specific antibody normalized to histone H3, with enrichment of the DUX4 promoter set to 1. (D) Immunoprecipitated chromatin was analyzed by qPCR using primers specific for DME1, DME2, the DUX4 promoter (Pro), the DUX4 transcription start site (TSS), or DUX4 intron 1 (Int1). Data are presented as fold enrichment of each target region by anti-Pol II relative to control IgG. For all panels, each bar represents the average of at least three independent ChIP experiments performed in triplicate.

DME1 and DME2 are in a permissive transcriptional state in most FSHD myocytes.

Active tissue-specific enhancers are generally devoid of CpG methylation and may also show lower nucleosome occupancy (49–51). To determine the DNA methylation status and nucleosome occupancy of DME1 and DME2 in FSHD myocytes, we performed NOMe-seq, a recently pioneered technique that has been instrumental in assessing specific chromatin states both for single loci and across the genome (39, 50–53). Taking advantage of the fact that mammalian cells lack GpC methylation, intact nuclei are incubated with the GpC methyltransferase M.CviPI, which methylates GpC dinucleotides not associated with nucleosomes or tightly bound transcription factors while leaving nucleosomal or transcription factor-occupied sites unmethylated. Following bisulfite conversion, patterns of GpC and CpG methylation are assessed, allowing the determination of both nucleosome occupancy and endogenous methylation status for individual DNA strands.

As expected, in differentiated 17Abic FSHD myocytes, we observed a well-defined nucleosome-depleted region (NDR) at the MyoD core enhancer in nearly all clones analyzed (Fig. 5A), similar to that reported for a rhabdomyosarcoma cell line, which also expresses MyoD (39). In contrast, DME1 and DME2 contained no well-defined NDR although several clones contained regions of local accessibility (Fig. 5B and C). We found GpC-methylated regions of 40 to 140 bp in 11 out of 17 DME1 modules and in 5 out of 16 DME2 modules, and two DME1 modules had a depletion of at least one nucleosome (Fig. 5B and C). Two of these accessible regions in DME1 overlap binding motifs for the looping factor CTCF as well as MyoD/myogenin and Runx1, which can recruit chromatin remodeling complexes to initiate gene expression (Fig. 5B) (54, 55). Assessment of endogenous CpG methylation across the assayed regions of DME1 and DME2 indicated that both enhancers are hypomethylated in the majority of clones (23% and 17% overall methylation, respectively) (Fig. 5B and C). We did not assess nucleosome occupancy and the methylation status of DME1/DME2 (DME1/2) in proliferating myoblasts or fibroblasts since these enhancers are not active with the DUX4 promoter in these cell types. Thus, in most myogenic cells, DME1 and DME2 appear to be associated with nucleosomes but not permanently silenced by DNA methylation. Taken together with our ChIP analysis, in which DME1 displays the marks of a bivalent enhancer and DME2 is enriched for Pol II occupancy, this finding indicates that these enhancers are in a permissive or poised state, consistent with the fact that DUX4 is expressed in only ∼1 out of 1,000 FSHD myonuclei (27).

FIG 5.

DME1 and DME2 are hypomethylated and occupied by nucleosomes in most FSHD myocytes. (A to C) Scheme depicting regions of the MyoD core enhancer (enh) (A), DME1 (B), and DME2 (C) analyzed by NOMe-seq (51). The MyoD core enhancer lies ∼20 kb upstream of the MyoD transcription start site (88). Positions of DME1 and DME2 are indicated as bp upstream of the DUX4 MAL start codon (27). Black circles represent CpG dinucleotides, and blue circles represent GpC dinucleotides. Gray fill indicates endogenous CpG methylation, and teal fill indicates exogenous GpC methylation (accessibility to M.CviPI). GCG trinucleotides that are methylated (marked with an X) are excluded from analysis since CpG versus GpC methylation cannot be distinguished in these cases. Pink bars represent regions of M.CviPI inaccessibility of at least 150 bp, indicating nucleosome occupancy. Regions where GpC methylation overlaps CTCF motifs, E-boxes, and Runx motifs are indicated.

DME1 and DME2 contain transcription factor binding motifs associated with early development and skeletal muscle gene expression.

DUX4 is thought to have originated following a gene conversion event in the DUXC macrosatellite array that occurred in the primate and Afrotheria lineages and subsequent translocation to 4qter in primates (56, 57). Restriction of DUX4 to these evolutionary lineages as well as the overall lack of synteny for chromosome 4q35 precludes a complete phylogenetic analysis of its cis regulatory regions to identify conserved motifs. Nonetheless, MatInspector analysis of DME1 and DME2 revealed clusters of transcription factor binding motifs commonly found in the regulatory regions of genes expressed in skeletal muscle and during embryogenesis, consistent with these regions functioning as myogenic or developmental enhancers (see Fig. S2 in the supplemental material).

DME1 contains a local enrichment of GC-rich E boxes [CA(C/G)(C/G)TG], the preferred binding sites for the muscle-regulatory factors MyoD and myogenin (OMIM 159970 and OMIM 159980, respectively), in heterodimers with their E-protein partners (see Fig. S2A in the supplemental material) (58–60). These transcription factors are required for the specification and differentiation of skeletal muscle, and their cognate binding motifs are highly concentrated in muscle genes expressed during differentiation (58, 61). One of these motifs in DME1 is a perfect match to the consensus sequence proposed for MyoD/E12 binding [(A/C/G)CA(C/G)(C/G)TGT(T/C)] (62). DME1 and DME2 also contain binding motifs for other key regulators of muscle gene expression, including Pax (paired and homeodomain motifs), TEAD, TEF-1 (MCAT), Six1/4 (MEF3), Meis, and Gli (the downstream effectors of Sonic hedgehog) (see Fig. S2A and B in the supplemental material). Many of these factors cooperate with each other or with muscle-specific cofactors to coordinately regulate the program of skeletal muscle differentiation (63–67). Both enhancers contain sequences that closely match the consensus binding site for CTCF, a transcription factor involved in the establishment and maintenance of higher-order chromatin architecture, which has been shown to cooperate with MyoD in the induction of myogenesis (68). Although ChIP assays for MyoD and myogenin showed no local enrichment at DME1 or DME2 (data not shown), our NOMe-seq assays indicated that these regions are occupied by nucleosomes in most cells. Thus, assessing transcription factor occupancy at these enhancers in the rare cells in which they are active may prove technically unfeasible. Although functional characterization of these motifs is beyond the scope of this study, the presence of these sequences supports our functional and interaction data indicating that DME1 and DME2 are myogenic enhancers of DUX4 transcription.

Methylation of DME1, DME2, and the DUX4 promoter ablates enhancer activity.

DNA hypomethylation of the D4Z4 repeat, and thus the DUX4 gene, is a hallmark of FSHD1 and FSHD2. To determine if DNA methylation levels at the DUX4 promoter affect the activity of DME1 and DME2, we tested the activity of in vitro-methylated DUX4 constructs (pJΔ, DME1-pJΔ, pJ, and DME2-pJ) in C2C12 myocytes. All constructs contain the DUX4 promoter present in the D4Z4 repeat. Methylation of all constructs with the CpG methyltransferase SssI ablated levels of DUX4-fl, confirming that the DUX4 promoter and both enhancers are inactive in the context of CpG methylation (see Fig. S3 in the supplemental material). Treatment with the HpaII methyltransferase, which methylates CpGs within CCGG motifs, decreased the activity of DME2 (5-fold reduction in activity of DME2-pJ versus only 2.7-fold decrease in activity of pJ), whereas it had no effect on the activity of DME1 (see Fig. S3). While HpaII sites are abundant in the DUX4 promoter, there is only one HpaII site in DME1 and one in DME2. Similarly, although DME1 and DME2 contain 11 and 12 CpG dinucleotides, respectively, the DUX4 promoter is ∼70% GC rich, with 86 CpGs within 1 kb upstream of the start codon. Thus, as expected, hypomethylation of the DUX4 promoter as found in FSHD is required for both enhancers to activate DUX4 expression.

Interactions of DME1 and DME2 with FSHD candidate genes.

The genomic region containing DME1 and DME2 is proximal to and not part of the invasive D4Z4 retroelement array, suggesting that these enhancers may have originally evolved to regulate other genes and were pathologically co-opted by DUX4. Since DME1 and DME2 are also positioned near two other FSHD candidate genes, FRG2 and FRG1 (OMIM 601278), we tested whether these enhancers physically interact with either of these genes in a cell- and disease-specific manner using 3C assays on differentiated FSHD (17Abic) and unaffected (17Ubic) myogenic cultures, as described above. In the absence of primary fibroblasts from donor family 17, we tested the muscle specificity of these interactions using primary human fibroblasts purchased from Lonza.

Interactions between DME1 and the FRG1/FRG2/DUX4 loci were interrogated using a series of PCR primers flanking BglII restriction sites (Fig. 6A) with the DIR1-containing BglII fragment as the anchor (primer 5) and normalized to a BAC control library. In FSHD myocytes, we observed a peak of interaction with FRG2 (primers 11 and 12) as well as a small interaction peak with DUX4 (primer 3) (Fig. 6B). These observations likely represent physical associations between DME1 and the promoters of FRG2 and DUX4. While the interaction between DME1 and DUX4 was present in all three cell types, the interaction between DME1 and FRG2 was much stronger in FSHD myocytes than in unaffected myocytes (as indicated by the size of the peak) and not present in fibroblasts (Fig. 6B). This suggests that the strong association between DME1 and FRG2 is specific to FSHD myocytes and is consistent with FRG2 mRNA being overexpressed in FSHD myocytes and increased during myogenic differentiation (21, 69). Surprisingly, we found that the relative interaction frequency in the region between primers 6 and 10 varies widely among the different cell types, suggesting a cell-type-specific genomic architecture in this region. No interaction between DME1 and FRG1 was observed in the myogenic cells (Fig. 6B, primers 15 to 16).

To interrogate interactions with DME2, the BglII fragment defined as DIR2 was chosen as the anchor fragment (Fig. 6A, primer 7). We observed a small interaction peak with FRG2, which was present in all three cell types (Fig. 6C, primers 11 and 12). The strong interaction peak with DUX4, also seen in Fig. 1, was present in both FSHD and unaffected myocytes but significantly reduced in fibroblasts, suggesting a muscle-specific association between DME2 and the DUX4 promoter (Fig. 6C, primer 3). No interaction between DME2 and FRG1 was observed in the myogenic cells (Fig. 6C, primers 15 and 16).

Because both enhancers associate with the DUX4 promoter in control as well as FSHD myocytes (i.e., the interactions were not disease specific), we did not repeat the 3C analysis in cells from additional donors. Taken together, our 3C, ChIP, NOMe-seq, and qRT-PCR data indicate that DME1 and DME2 are functional enhancers in differentiated skeletal myocytes. In the majority of FSHD myocytes, these enhancers are hypomethylated but associated with nucleosomes and likely not interacting with any gene promoters. In rare cells, these permissive enhancers become activated and associate with the hypomethylated DUX4 promoter to drive expression of DUX4-fl in skeletal muscle.

DISCUSSION

We have described the identification of two enhancers upstream of the 4q D4Z4 array that drive DUX4-fl expression in differentiated skeletal myocytes and may be implicated in the predominantly muscle phenotype of FSHD. This is the first identification of enhancers capable of interacting with the DUX4 promoter and activating DUX4 expression. D4Z4 repeats have been described as containing an enhancer element that activates the FRG1, FRG2, and DUXC promoters but has no effect on the DUX4 promoter (26, 70). Importantly, although DME1 and DME2 are enriched for enhancer marks in several cell types (http://genome.ucsc.edu/cgi-bin/hgGateway) and may regulate other genes in nonmuscle cells, they are active with the DUX4 promoter only in differentiated skeletal myocytes and not fibroblasts. Additionally, the activities of these enhancers are dependent upon an epigenetically permissive DUX4 promoter. This finding is consistent with the DUX4 expression observed in FSHD1 and FSHD2 myocytes, where the DUX4 promoter is hypomethylated, and with the extremely rare expression in unaffected myocytes or nonmuscle cells, where the DUX4 promoter is methylated (27, 28; also T. I. Jones and C. L. Himeda, unpublished data). Thus, DUX4 promoter methylation is one mechanism by which the activities of DME1 and DME2 can be modulated, potentially resulting in the variable expression of DUX4-fl and, consequently, variable onset and severity of FSHD pathology.

Although the pathogenic expression of DUX4 in FSHD occurs in skeletal muscle, DUX4 is expressed normally in testis and in pluripotent stem cells (27), and a recent report describes expression of DUX4-fl in nonmuscle somatic tissues of both FSHD and normal fetuses (45). Recently, the cell adhesion molecule FAT1 (OMIM 600976), a downstream target of DUX4, has been described as a potential FSHD modifier during development (71). Consistent with this possibility, the predominance of developmental binding motifs (Sox, Runx, forkhead, Smad, and homeobox motifs) in DME1 and DME2 suggests that these enhancers have normal gene targets in multiple lineages during development. The Pax family, for example, recognizes paired/homeodomain motifs in its target genes and plays an essential role in organogenesis (72). Two members of this family, Pax3 and Pax7, are expressed in embryonic muscle progenitors, and Pax7 is required for the specification and maintenance of muscle satellite cells (73). It has been suggested that aberrant DUX4 expression in satellite cells might lead to a progressive loss of muscle regenerative capacity over time (32), and it is possible that Pax3 and Pax7 contribute to this by activating DME1 and DME2 in muscle progenitors. Thus, the primate-specific insertion of the DUX4 retrogene directly downstream of these enhancers may have pathological consequences at very early stages in FSHD patients that do not manifest until much later in life.

The presence of CTCF motifs in DME1, DME2, and the D4Z4 repeats is intriguing and hints at a possible mechanism for the establishment of looping interactions between these proximal enhancers and the DUX4 promoter. It has recently been demonstrated that the majority of long-range genomic interactions overlap but are not blocked by regions of CTCF enrichment, indicating that while this factor likely plays a key role in the establishment of enhancer-promoter interactions, it does not always demarcate physically insulated chromatin domains (74). The D4Z4 arrays on 4q and 10q show similar patterns of specific histone alterations in FSHD cells, suggesting that these regions are conjointly deregulated (15). It will be important to determine if CTCF plays a role in bridging interactions between the D4Z4 repeats on 4q and 10q and between DME1/2 and distant loci. Many well-characterized enhancers are known to regulate the expression of far-distant genes (e.g., the limb enhancer of Sonic hedgehog lies 1 Mb away from its target promoter), and recent genome-wide studies have greatly expanded the identification of distal regulatory regions (74). Interestingly, two FSHD1 patients with large deletions in the D4Z4 proximal region have been described (75). Although these deletions encompass DME1 and DME2, this region is still intact on the other 4q allele and on chromosome 10. Thus, although interactions in trans are likely to be rare events, it is conceivable that these enhancers might function in trans to drive DUX4-fl in these unique cases, keeping in mind that developing FSHD pathology requires only a few myogenic cells to express DUX4-fl (76). The enrichment of RNA Pol II, particularly at DME2, suggests that these enhancers may be templates for the production of ncRNAs, which might also play a role in regulating DUX4 or other genes (77).

DME1 and DME2 are uniquely situated to drive the pathological misexpression of DUX4 in FSHD myocytes. Taken together, our data indicate that interactions between both enhancers and the DUX4 promoter might well drive sporadic bursts of DUX4 expression in rare muscle nuclei. Transcriptional bursting has been observed as a widespread phenomenon in organisms from bacteria to mammals, especially for genes with low-activity promoters, and is thought to be dependent on promoter architecture (78). Although the architecture and kinetics of the DUX4 promoter have yet to be characterized, it is reasonable to assume that dynamic fluctuations in nucleosome and transcription factor occupancy at both the promoter and enhancers of DUX4 may result in bursts of toxic DUX4-fl expression (Fig. 7C). Importantly, while DME1 is associated with DUX4 in fibroblasts as well as myocytes, the enhancer is only active with the DUX4 promoter in myocytes. Both DME1 and DME2 interact with DUX4 in normal as well as FSHD myocytes; however, the DUX4 promoter is present in multiple copies on the chromosome (one per D4Z4 unit), and unaffected individuals have longer arrays. While it is technically unfeasible to assess interactions specific to the distal DUX4 promoter, these enhancers are likely to be competitively engaged with DUX4 promoters throughout the array, with promoters at the proximal and distal ends of the array being more accessible. Therefore, contraction of the array and/or epigenetic derepression of the D4Z4 region seen in FSHD should increase the likelihood of engagement with the distal promoter, resulting in production of the stable, pathogenic transcript (Fig. 7). Engagement with the proximal promoter might also lead to DUX4-fl expression since cryptic promoters are capable of driving expression from nonadjacent genes (79).

FIG 7.

Model of DME1 and DME2 driving DUX4-fl expression in FSHD myocytes. (A) Schematic diagram of physical interactions between DME1/2 and the promoters of FRG2, DUX4 (distal unit, black triangle), and other potential gene targets. The thickness of each line denotes the strength of the interaction, as seen in 3C assays (Fig. 6). (B) Schematized drawing depicting the 4qA subtelomere from an unaffected and an FSHD1 individual, each of whom has a permissive allele for DUX4-fl expression (encoding a polyadenylation signal in exon 3). The DUX4 gene is contained within every unit (triangle) of the D4Z4 repeat array. FSHD1 patients have a contracted D4Z4 array (1 to 10 units) on one allele, whereas unaffected subjects have a normal-sized array (11 to 150 units) on both alleles. In unaffected individuals, the D4Z4 array is under epigenetic repression (gray oval), which is lost in FSHD1. In both skeletal myocytes and fibroblasts from FSHD1 patients, proximal enhancers DME1 and DME2 are most likely competitively engaged with DUX4 promoters throughout the array. Due to repeat contraction and loss of epigenetic silencing in FSHD1 cells, these enhancers have a higher frequency of association with the distal DUX4 promoter (black triangle). However, DME1 and DME2 are active only with the DUX4 promoter in myocytes; thus, interactions with the distal DUX4 promoter drive expression of the pathogenic DUX4-fl transcript in myocytes but not fibroblasts. (C) Potential states of the DMEs are represented. Repressed enhancers have methylated DNA and are occupied by nucleosomes with repressive modifications; poised enhancers have hypomethylated DNA and remain occupied by nucleosomes which display active or bivalent modifications; active enhancers are depleted of positioned nucleosomes and interact with transcription factors (TxF) and their target promoters. In differentiated myocytes, DME1 and DME2 are predominantly in the poised state and sporadically in the active state, interacting stochastically with the distal DUX4 promoter and regulating bursts of DUX4-fl expression. Tel, telomere.

The low frequency of association between DME1 and DUX4 may be due in part to competition with other gene promoters, including the FRG2 promoter. Our finding that the association between DME1 and the FRG2 promoter is much stronger in FSHD than in unaffected myocytes is consistent with reports that FRG2 is expressed in differentiating FSHD myoblasts, whereas expression in normal skeletal muscles ranges from very low to undetectable (21, 69). In contrast, DME2 shows a very strong and preferential engagement with the DUX4 promoter over the FRG2 promoter, and this association appears to be muscle specific. This is particularly noteworthy in light of a recent report describing the first D4Z4-DUX4 transgenic mouse model (80). While the target mouse (D4Z4-2.5) in this study is transgenic for an FSHD1 allele containing DME1 but not DME2, the control mouse (D4Z4-12.5) is transgenic for an allele containing both enhancers. Despite hypermethylation of the longer D4Z4 array in the D4Z4-12.5 mouse, which appears to silence expression in most nonmuscle tissues, DUX4 transcripts are still detected consistently and reproducibly in two anatomical muscles. Thus, a transgenic mouse containing DME2, which we show interacts strongly with the DUX4 promoter in both FSHD and unaffected myocytes but not in fibroblasts, displays largely muscle-restricted expression of DUX4. Based on these combined findings, we propose that DME2 is important for the expression of DUX4 in FSHD muscle and also for the rare DUX4 expression detected in muscles from unaffected individuals (28).

It is worth pointing out that a combination of techniques assessing histone modifications, chromatin accessibility, nucleosome occupancy, DNA methylation, and looping interactions is required to gain a realistic picture of enhancer activity. Each of these techniques comes with its own set of limitations and, especially in the case of these unusual DUX4 enhancers, can easily lead to very different interpretations if performed in isolation. Our data utilizing multiple assays, confirmed with the independent ENCODE data, are consistent with a model in which DME1 and DME2 are in a permissive state in most cells. These enhancers are associated with nucleosomes, but they are not permanently silenced by DNA methylation, and they contain marks of bivalent or poised enhancers. In rare FSHD myocytes, in response to unknown signals, a pioneer transcription factor(s) such as MyoD or Runx1 might outcompete nucleosomes for binding to DME1/2, resulting in chromatin remodeling, association with the hypomethylated DUX4 promoter, and pathological expression of DUX4-fl.

It will be important to determine which physiological stimuli and downstream pathways regulate the activities of DME1 and DME2. Although primarily a disorder of skeletal muscle, FSHD is also associated with retinal vasculopathy, high-tone deafness, and intellectual disabilities (81–83), which could reflect environmental cues triggering common networks of transcription factors in different tissues. In addition to the motifs described above, both enhancers contain putative binding sites for chromatin modifiers and hormone-responsive factors which might regulate activity in diverse cell types. Wnt signaling has recently been shown to negatively regulate DUX4 expression; inhibition of the canonical Wnt pathway leads to enhanced DUX4 expression in cultured FSHD myotubes, yet no putative binding sites for Wnt effector proteins exist within the D4Z4 array (84). DME1 and DME2 do not contain motifs for Wnt effectors; however, we report motifs matching a T-cell factor (TCF)/LEF consensus (YYCTTTGWW) (85) elsewhere in DIR1 and DIR2, in addition to a site upstream of the diagnostic p13-E11 sequence, all of which are candidates for direct regulation of DUX4 by Wnts. It is worth emphasizing that only a small fraction of FSHD myonuclei express DUX4-FL that is detectable by immunocytochemistry (27, 28), and patients have many relatively unaffected muscles. Binding of the downstream effectors of dynamic signaling pathways to DME1 and DME2 may represent another mechanism by which DUX4 levels are modulated, in addition to the proposed regulation by antisense transcripts (86). With the majority of disease- and trait-associated genetic variants concentrated in noncoding DNA (87), it seems likely that polymorphisms within DME1 and DME2 might also play a role in the variable DUX4-fl expression among FSHD patients (28), as well as in the clinical variability, ranging from nonmanifesting to clinically severe disease, seen in different genetically diagnosed FSHD1 individuals.

Beyond genetic alterations, epigenetic variability among families is likely to be a major determinant of penetrance and severity in FSHD. SMCHD1, a modifier of FSHD1 and causative lesion in FSHD2 (18, 20), affects DNA methylation of the D4Z4 arrays, and different families display differential methylation of the DUX4 promoter and gene body (Jones and Himeda, unpublished). Determining the genetic and epigenetic mechanisms by which DUX4 expression is regulated and the FSHD phenotype is established remains a significant challenge for future studies.