Igf2 adult-specific skeletal muscle enhancer activity revealed in mice with intergenic CTCF boundary deletion

Abstract

Precise, monoallelic expression of imprinted genes is governed by cis regulatory elements called imprinting control regions (ICRs) and enhancer-promoter (E-P) interactions shaped by local chromatin architecture. The Igf2/H19 locus employs allele-specific CTCF binding at the ICR to instruct enhancer accessibility to maternal H19 and paternal Igf2 promoters. Here, we investigate the CTCF-bound centrally conserved domain (CCD), intergenic to H19 and Igf2, and an adjacent widely expressed lncRNA. Using transgenic mice, deletion alleles reinforced CCD as a neonatal muscle-specific repressor of maternal Igf2. However, deletion of the abutting lncRNA did not affect Igf2 levels. Unexpectedly, in adult skeletal muscle where Igf2 is normally repressed, absence of CCD resulted in remarkable, high-level activation of Igf2 from both parental alleles. Through multimodal chromatin analyses, we identified a conserved putative adult skeletal muscle enhancer (PaSME) insulated between chromatin domains at ICR and CCD. We propose that removal of CCD allows PaSME to drive robust abnormal Igf2 activation on both alleles in adult skeletal muscle. Thus, we uncover CCD as a developmental biallelic muscle-specific repressor, adding a new layer of architectural regulation to the extensively studied Igf2/H19 locus.

Author summary

All mammals have a unique class of genes that are crucial for normal growth, known as “imprinted genes,” where only one gene copy inherited from a parent is active. Imprinted genes are usually expressed in specific cell types and at certain stages of development, but the instructions for how this pattern is achieved is not fully understood. Our study uses the well-known imprinted Igf2/H19 locus as a model for studying complex gene regulation, as Igf2 plays important roles in muscle cell growth. Between the two genes lies a genetic boundary called the Centrally Conserved Domain (CCD) and a proposed activating sequence that we called the putative adult skeletal muscle enhancer (PaSME). Using transgenic mice and advanced DNA folding analyses, we show that CCD acts as a blocker of PaSME. This barrier keeps Igf2 expression appropriately off in adult muscle. When CCD is deleted, this barrier is lost, allowing PaSME to activate Igf2 to abnormally highly levels in muscle, leading to increased body weight in some animals. This mechanism enhances our understanding of how imprinted genes are controlled at critical timepoints, and their impact on development and disease.

Introduction

Imprinted genes comprise a small but developmentally important set of genes in mammals that are expressed from a single parental allele. Imprinted gene regulation requires cis elements, including imprinting control regions (ICRs) and enhancers, to achieve allele and tissue specific expression essential for normal development [1]. At the imprinted murine Igf2/H19 locus, the paternally-expressed insulin growth factor 2 (Igf2) gene and the maternally-expressed long non-coding RNA (lncRNA)-encoding H19 gene are regulated by the intergenic 2 kb ICR (2kb upstream of H19) in combination with enhancers downstream of H19 and Igf2. The mouse and human IGF2/H19 loci share spatial and sequence conservation of regulatory elements [2]. The ICR serves as a DNA methylation-sensitive CTCF boundary, mutations and epimutations of which are associated with imprinting disorders Beckwith-Wiedemann syndrome and Silver-Russell syndrome [2,3]. Although tissue-specific enhancers for embryonic and neonatal liver and skeletal muscle have been defined [4–6], the ubiquitous expression patterns of Igf2 and H19 across development require additional undefined cis-acting sequences.

Imprinted Igf2 and H19 levels increase throughout preimplantation development in all tissues and are repressed postnatally, with a few exceptions. For example, in adult skeletal muscle, H19 expression persists while Igf2 is repressed [7–9]. In adult brain, Igf2 persists and is biallelically expressed in specific tissues such as choroid plexus and leptomeninges, or is maternally-expressed in a subset of neurons [7,10–14]. Together, these exceptions indicate that distinct regulatory mechanisms direct H19 and Igf2 expression in brain and adult skeletal muscle.

To address other potential regulatory elements that govern tissue and developmental-specific expression, we and others examined the 130-kb region spanning the Igf2/H19 locus in vivo. A conserved 2kb GC-rich region was identified between H19 and Igf2 that is hypomethylated and contains prominent and ubiquitous DNase I hypersensitive sites (DHS) [15]. This region, coined the centrally conserved domain (CCD) [16], is composed of two elements with high identity to human sequence [17]. Unlike the maternally-bound ICR, CTCF and cohesin were detected on both alleles at CCD [18]. Furthermore, transgenic and deletion studies evaluating CCD indicated that it may function as an Igf2 brain-specific enhancer and/or a skeletal muscle-specific repressor of the maternal Igf2 allele, likely through CTCF-mediated looping [16,17,19,20].

Adjacent to CCD lies a lncRNA (designated intervening region or Ivr) identified in embryonic and neonatal tissues with hallmarks common to most lncRNAs (RNA polymerase II transcribed, polyadenylated and spliced) [21,22]. Transcription of Ivr was reported to be in the direction of and anti-sense to Igf2, but the role of Ivr at the imprinted H19/Igf2 locus is unknown. Given the coincident expression of Ivr and Igf2 in early development, and the proximity of Ivr to CCD, we hypothesized that Ivr may contribute to CCD-directed tissue-specific regulation of Igf2.

Here we further analyze lncRNA Ivr and its association with CCD-dependent Igf2 expression at different stages in development. First, we demonstrate that Ivr is preferentially expressed from the paternal allele and its paternal-specific expression is dependent upon the Igf2/H19 ICR in multiple tissues. Next, by deleting Ivr, we show that it is not required for CCD-dependent repression of maternal Igf2 in skeletal muscle. Co-deletion of Ivr and CCD recapitulates reported skeletal muscle-specific activation of maternal Igf2 in neonates [17,19]. Moreover, in adults, deletion of CCD and Ivr from either allele results in high level Igf2 expression in skeletal muscle. Our 3D chromatin analyses in skeletal muscle, together with publicly available 2D chromatin data, indicate the CCD interaction with sequence proximal to Igf2 serves as a boundary to adult-specific skeletal muscle enhancer activity. The biological function of Ivr remains to be elucidated.

Results

Characterization of the intervening region (Ivr) transcript adjacent to the Igf2/H19 intergenic CCD-hypersensitivity site

To identify additional regulatory elements that may contribute to tissue- and allele-specific Igf2 expression, we examined the transcript (Ivr) adjacent to CCD, which was first recognized in a mouse imprinting region tiling array (MIRTA) that described lncRNAs associated with imprinted loci [22]. Ivr was characterized as a spliced, ~ 8kb transcript expressed in embryonic and neonatal tissues (Fig 1A annotation and 1B schematic). Two Ivr transcription start sites (TSSs) and Exon1/2 splice junctions were identified from alignment of cloned 5’RACE DNA (Sheet A in S1 Table) [22]. Embryonic samples from the Mouse ENCODE project showed transcription coincident with Ivr in the direction of and anti-sense to Igf2 (Fig 1A and Sheet A in S2 Table). Publicly available DHS and ChIP-seq data from multiple embryonic tissues indicated that the Ivr TSSs are adjacent to the DHS and CTCF binding sites of CCD. Additionally, the Ivr promoter is positioned within open chromatin decorated with signatures of promoters as well as enhancers (Pol2, H3K4me1, H3K4me3 and H3K27ac) (Fig 1A and Sheet A in S2 Table) [23–25].

Fig 1. Analysis of Igf2/H19 intervening lncRNA Ivr and CCD.

(A) UCSC genome browser view of annotated CCD and Ivr regions using publicly available ENCODE data, including tissue-specific RNA-seq, DNAse HS-seq and ChIP-seq for relevant protein and histone marks (Sheet A in S2 Table). (B) Schematic for CCD (purple box)/Ivr (gray boxes) with relative locations of primers used to identify transcription start sites (TSSs) and splice junctions (r, s), primers used for allelic expression (a, b) and SNP (*) (Sheet C in S3 Table). AvrII digestion discriminates the C57BL/6 (B) and Mus Castaneus (C) Ivr transcripts. Allele-specific expression for H19 (pink box), Igf2 (blue box) and Ivr are depicted for WT alleles and ΔICR alleles. Tissue-specific enhancers (green circles) regulate maternal H19 and paternal Igf2 (curved arrows) and may regulate paternal Ivr (dashed arrow). CTCF bound ICR (yellow box) blocks enhancers from engaging maternal Igf2. (C) Allele-specific Ivr in WT (BxC and CxB) neonatal liver (Li). (D) Allele-specific Ivr from WT and ΔICR alleles in neonatal Li, heart (H) and skeletal muscle (SkM). Primer pairs ‘a’ and ‘b’ in (B) were used in (C) and (D), respectively. For (C) and (D), genotype and ladder lane M (BR322 DNA-MspI) are above panels; % expression from paternal allele [%p = 100xp/(p + m)] of AvrII digested qRT-PCR DNA are noted below panels. U corresponds to undigested PCR product. (E) Maternal WT versus ΔICR Ivr expression [m/(m + p)] in neonatal (Neo) Li and SkM. (F) Paternal WT versus ΔICR Ivr expression [p/(m + p)] in Neo Li. (Two-tailed Welch’s T-test, ***p < 0.001; ****p < 0.0001; error bars represent SD).

We evaluated the allele-specific expression of Ivr because it was intervening to imprinted genes H19 and Igf2. Using RT-PCR with primers in Exons 2 and 3, we first confirmed expression of Ivr in neonatal tissues and sequenced Exon 2/3 splice junction from cDNA (Fig 1B, 1C, and 1D and Sheet B in S1 Table). Next, in progeny from reciprocal matings between C57BL/6 (B) and Cast7 (C) mice, which have a Mus castaneus chromosome 7 on a B background, we evaluated Ivr allelic expression by restriction fragment length polymorphism (RFLP). Coincident with its location on the Igf2 side of the ICR, Ivr was preferentially expressed from the paternal allele in neonatal liver, skeletal muscle and heart (Fig 1B,1C,1D). Remarkably, inheriting the Igf2/H19 ICR deletion (ΔICR) [26] from the maternal allele resulted in higher maternal Ivr levels in skeletal muscle and heart but not liver (Fig 1A, 1D, 1E). However, on the paternal ΔICR allele, Ivr levels were reduced in liver (Fig 1D, 1F). These results follow the trend for Igf2 tissue-specific activation on the ΔICR maternal allele (high in skeletal muscle and heart, low in liver) and repression on the ΔICR paternal allele (most repressed in liver) [26,27]. The tissue-specific Igf2 levels from ΔICR alleles may be due to how each tissue-specific enhancer is shared by H19 and Igf2 on the ΔICR alleles. Our results suggest that Ivr expression is partly regulated by tissue-specific Igf2 shared enhancers downstream of H19 on the paternal allele that are blocked by ICR insulator activity on the maternal allele. While the skeletal muscle enhancer (SME) and endodermal enhancer (EE) for liver have been identified, heart specific enhancer remains undefined (S1 Fig genome browser annotation) [4,5].

Generation and analysis of ΔIvr and ΔCCDIvr alleles

As stated above, transgenic and deletion studies evaluating CCD indicated that it may be a skeletal muscle-specific repressor of the maternal Igf2 allele [17,19]. To test the hypothesis that Ivr is required for CCD to repress muscle-specific Igf2 expression, we derived several deletion alleles in mice. We generated an allele with Ivr deleted and CCD floxed (S2A Fig). Mice with the targeted ΔIvr + Neo allele were mated with Flp mice for removal of NeoR cassette to generate the ΔIvr allele. These offspring were then mated with Cre mice to generate mice with the ΔCCDIvr allele. Alleles were confirmed by Southern blot and PCR analyses (S2A and S2B Fig and Sheets A and B in S3 Table). The ΔIvr and ΔCCDIvr mice were mated with the Cast7 mice to generate progeny for Igf2 allelic expression.

Igf2 is derepressed from the maternal ΔCCDIvr but not the ΔIvr allele in neonatal skeletal muscle

In neonates inheriting the maternal ΔIvr allele, maternal Igf2 was not activated (Fig 2A), demonstrating that the Ivr alone does not repress maternal Igf2. As expected, neonates inheriting the maternal ΔCCDIvr allele exhibited maternal Igf2 activation exclusively in neonatal skeletal muscle (tongue, Fig 2B and 2C) [17,19]. In contrast, Igf2 levels were unchanged in neonates inheriting the paternal ΔCCDIvr allele (Fig 2B; S2C Fig). Furthermore, H19 was unaffected regardless of the transmission of the ΔCCDIvr alleles (S2D, S2E Fig). Finally, we detected similar levels of Igf2 activation from the maternal ΔCCDIvr allele in neonatal skeletal muscle from tongue and hindlimb (Fig 2B, 2D). Our results for the ΔCCDIvr allele are consistent with the previous reports that (1) deleted the conserved CCD DHS sequence (leaving Ivr intact) from a 130-kb imprinted Igf2-H19 transgene [17] and (2) deleted the endogenous CCD together with Ivr sequence [19].Together, these data indicate that CCD is an Igf2 skeletal muscle-specific repressor on the maternal allele (Fig 2E). Despite the critical role of Igf2 in early growth, inheritance of the ΔCCDIvr alleles did not affect total body weight of neonates (S2F Fig), although there was a trend in weight gain in adult CCDIvr^Δ/+ and CCDIvr^+/Δ mice relative to wild-type (WT) littermates (S3A Fig). Because adult-specific weight gain was previously reported in mice expressing high levels of Igf2 in adult tissues [28], we next tested whether the observed trend in adult weight gain in ΔCCDIvr mice was associated with derepression of Igf2 in skeletal muscle.

Fig 2. Igf2 activation from the maternal ΔCCDIvr allele in skeletal muscle.

(A) Allele-specific Igf2 expression in liver (Li) and tongue (T) from neonates inheriting WT and ΔIvr allele paternally (CxB and CxΔIvr) and maternally (BxC and ΔIvrxC). For (A-D), Igf2 qRT-PCR products were digested with MluCI to discriminate the parental B (WT B, ΔIvr, ΔCCDIvr) and C alleles. In (B), allele-specific Igf2 expression in Li, T and heart (H) from neonates inheriting WT and ΔCCDIvr allele maternally (BxC and ΔCCDIvrxC) and paternally (CxB and CxΔCCDIvr). In (B) (T panel) and (D), Igf2 expression from maternal ΔCCDIvr allele is noted below the gel. (C) Igf2 expression from maternal WT versus ΔCCDIvr allele in neonatal T (Two-tailed Welch’s T-test, ****p < 0.0001; error bars represent SD.) (D) Maternal-specific Igf2 expression in adult TA, GAS, QUAD, T and choroid plexus, and neonatal skeletal muscle (hindlimb and T) WT (BxC) and mutant (ΔCCDIvrxC) tissues as indicated above gel. For panels (A), (B) and (D): M (ladder), U (undigested), B and C (control tissues), %m (%maternal = 100 x m/(m + p). (E) Schematic of H19, Ivr and Igf2 expression from WT, ΔIvr and ΔCCDIvr maternal alleles. Igf2 is expressed from the maternal ΔCCDIvr allele in neonatal SkM.

Igf2 is highly activated from both parental ΔCCDIvr alleles in adult skeletal muscle

We first evaluated allele-specific Igf2 expression by RFLP in tongue (T) and easily accessible larger hind limb muscles including gastrocnemius (GAS), quadriceps (QUAD) and tibialis anterior (TA) in adult mice inheriting the maternal ΔCCDIvr allele. Unexpectedly, we measured a large increase in maternal Igf2 activation in the adult hind limb muscles compared to the adult T (Fig 2D). We reasoned that this difference was linked to the extensive transcriptional diversity across muscle subtypes [29]. We therefore extended our analysis to include hind limb muscles extensor digitorum (EDL) (with an overlapping transcriptional profile to GAS, QUAD and TA) and soleus (SOL) (with an overlapping transcriptional profile to T) to capture other possible skeletal muscle-specific outcomes with the genetic deletion of CCDIvr. Accordingly, in adult CCDIvr^Δ/+ mice, maternal Igf2 was modestly activated in T and SOL (34% and 47% of total Igf2, respectively) (Fig 3A), similar to neonatal skeletal muscle (Fig 2B-2D). In stark contrast, maternal Igf2 activation in EDL, QUAD, TA and GAS greatly exceeded expression from the paternal WT Igf2 allele. Compared to WT, Igf2 was elevated on average 38-, 54-, 27- and 27-fold in CCDIvr^Δ/+ EDL, QUAD, TA and GAS, respectively (Fig 3B, 3C). These results suggest that maternal CCDIvr functions as an insulator, deletion of which allows maternal Igf2 access to robust skeletal muscle-specific enhancer activity. Note that as with neonatal tissues, Igf2 activation was not dependent upon Ivr alone, because Igf2 was not expressed from the maternal ΔIvr allele (Ivr^-/+) in adult TA (S3B Fig).

Fig 3. Igf2 expression and Capture C interactions with proximal Igf2 CTCF binding site in adult ΔCCDIvr skeletal muscle.

(A) Maternal allele-specific Igf2 expression in CCDIvr^+/+ versus CCDIvr^Δ/+ tongue (T) and soleus (SOL). (B) Total Igf2 expression in CCDIvr^+/+ versus CCDIvr^Δ/+ T, SOL, EDL, QUAD, TA and GAS. (C) Schematic of H19, Ivr and Igf2 expression from WT, ΔIvr and ΔCCDIvr maternal alleles (top) and paternal alleles (bottom) in adult skeletal muscle. (D) Total Igf2 expression in CCDIvr^+/+ versus CCDIvr^+/Δ T, SOL, EDL, QUAD, TA and GAS. For (B) and (D), relative Igf2 expression in mutant vs WT muscle normalized to 1, depicted on log10 scale (y-axis) (Two-tailed Welch’s t-test: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001; error bars represent SD). (E) Normalized Capture C tracks from CTCF sites at CCD (Anchor at CCD, not allele-specific) and sequence proximal to Igf2 (Anchor at Igf2prox, allele-specific) in adult skeletal muscle of WT, CCDIvr^Δ/+ (MKO) and CCDIvr^+/Δ (PKO) mice. All data presented on WashU Epigenome Browser using mouse mm10 (Sheet A in S2 Table). Interactions at CCD anchor (black - both alleles) and at Igf2prox anchor (pink - maternal alleles, blue - paternal alleles), are highlighted (purple bars). For CCD Anchor, WT-Adult presents interactions for both parental WT alleles, MKO-Adult presents track from WT paternal allele and PKO-Adult presents track from WT maternal allele. From Igf2prox anchor, on WT-Adult paternal versus PKO_Adult paternal, boxed regions indicated gain and loss of interactions to Mrpl23/SME CTCF sites and CCD, respectively. Wald statistics using two methods to identify peaks are presented (* or # p < 0.05, ** or ## p < 0.01, Sheet E in S5 Table).

We next examined Igf2 on the paternal ΔCCDIvr allele in adult muscle tissues. Chromatin conformation capture (3C) in human culture cells indicated that hypomethylated CCD sequence interacted with CTCF-bound sequence upstream of IGF2 on both alleles and included IGF2 in a chromatin loop or topologically-associated domain (TAD) [18,30]. We anticipated that this CTCF-dependent architecture could be blocking Igf2 from a regulatory sequence such as an enhancer in skeletal muscle. Thus, we hypothesized that for maternal and paternal ΔCCDIvr alleles, loss of CTCF anchor points in the sub-TAD would release the putative enhancer from insulation. This element would then be equally accessible to both parental Igf2 promoters and lead to similar Igf2 activation. Compared to WT mice, Igf2 levels were elevated on average 57-, 54-, 79-, 135-, 206- and 173-fold in CCDIvr^+/Δ T, SOL, EDL, QUAD, TA and GAS, respectively (Fig 3C, 3D). Unexpectedly, these Igf2 levels from the paternal ΔCCDIvr allele were much higher than Igf2 activation from the maternal ΔCCDIvr allele (Fig 3A, 3B). This result was in stark contrast to CCDIvr^+/Δ neonatal T in which paternal Igf2 expression was unaffected (Fig 2B; S2C Fig), supporting the role of a developmental and tissue-specific enhancer that is insulated by CCD. Furthermore, this enhancer activity was elevated by access to SME on the paternal ΔCCDIvr allele (Fig 3C, 3D).

Endodermal and skeletal muscle enhancers distal/centromeric to H19 have been characterized for neonatal H19 and Igf2 expression [4,5]. On the paternal ΔCCDIvr allele, Igf2 has access to skeletal muscle enhancer (SME) overlapping Nctc1 (S1A Fig) [5,31]. SME, which is blocked by the ICR on maternal allele, regulates paternal-specific expression of Igf2 and maternal specific expression of H19 in skeletal muscle. At least two scenarios could explain high Igf2 levels on the ΔCCDIvr alleles: (1) CCD normally blocks this SME from engaging Igf2 promoters and deleting CCD increases its probability to engage these promoters or (2) a previously uncharacterized SME between the ICR and CCD was exposed with the CCDIvr deletion (Fig 3C). To address these possibilities, we next investigated how loss of CCD-dependent architecture in neonatal and adult skeletal muscle tissues could account for elevated Igf2 from ΔCCDIvr alleles.

CCD interacts with CTCF-bound sequence upstream of Igf2 in neonatal and adult skeletal muscle on both alleles

The region surrounding the Igf2/H19 locus contains multiple CTCF binding sites that play a role in shaping local chromatin architecture. We first aimed to map the constitutive topologically-associated domains (TADs) within the Igf2/H19 imprinted locus. To this end, we utilized published and publicly available ENCODE data across multiple cell-types to confirm local TAD boundaries and associated CTCF/cohesin binding sites surrounding Igf2, including CCD (S1 Fig and Sheet A in S2 Table). Once the bulk of the chromatin landscape was defined, we then genetically dissected allele-specific CCD-dependent chromatin architecture in vivo using WT and ΔCCDIvr deletion alleles in mice. Using a targeted chromatin conformation approach, we interrogated the local Igf2/H19 gene architecture and reinforced the insulator role of CCDIvr in neonatal and adult skeletal muscle tissues.

We first retrieved Micro-C data performed in embryonic stem cells (ESCs) [32] to confirm the constitutive local TAD that encompasses the Igf2/H19 locus. We then aligned the TAD boundaries with CTCF (d0 Heart) and cohesin (CH12 cell line) ChIP sequencing data, as well as CTCF binding site motifs and motif orientation (S1 Fig and Sheets A and D in S2 Table). This analysis revealed multiple intra-TAD intersecting ‘dots’ between convergent CTCF sites (S1 Fig, centromeric green CTCF and telomeric red CTCF), indicating “pause sites” for loop extrusion [33–35]. ‘Dots’/loops between Igf2prox CTCF sites and multiple convergent CTCF sites across the locus are marked in S1 Fig (purple arrowheads), and include CTCF bound CCD (S1 Fig, dashed lines), supporting our hypothesis that CCD is contributing to local gene architecture. These findings were consistent with chromatin conformation analysis in human cell lines demonstrating that CCD interacts with CTCF-bound sequence proximal to IGF2 (syntenic to Igf2prox, S1 Fig and Sheet D in S2 Table) on both parental alleles [30]. In mouse cell lines, this biparental CCD/Igf2 interaction had not been demonstrated. Previous work using allele-specific CTCF-ChIP and 4C-seq chromatin conformation analysis in hybrid mouse ESCs mapped allele-specific architecture surrounding the ICR and Igf2prox CTCF binding sites at the H19/Igf2 locus, but CCD was not addressed (S1 Fig) [34]. We also highlighted multiple looping/intra-TAD interactions from the Igf2prox CTCF sites, underscoring the need to consider additional cis-regulatory elements as important regulators of Igf2/H19 expression (Micro-C in S1 Fig) [32–34,36].

To assess whether the previously characterized chromatin architecture in pluripotent cell line is preserved in ΔCCDIvr skeletal muscle tissues with elevated levels of Igf2 expression, we conducted allelic chromatin conformation analyses in neonatal and adult F1 hybrid maternal and paternal ΔCCDIvr skeletal muscle nuclei (Sheets A-C in S4 Table). Using next-generation Capture-C technology, we designed probes to capture interactions from predicted looping sites across the TAD encompassing the Igf2/H19 locus [37]. Where available, SNPs in the hybrid mice enabled discrimination of maternal versus paternal allele-captured interactions (S4 Table). We first examined allele-specific chromatin interactions on WT alleles in neonatal (WT-Neo) and adult (WT-Adult) skeletal muscle (S1 Fig, anchors at CCD [no SNP], Igf2prox [SNP], and ICR [SNP]) and confirmed known looping interactions of Igf2prox (paternal allele) and ICR (maternal allele) to convergent CTCF sites centromeric to H19 in skeletal muscle, as demonstrated in ESCs [34] (S1 Fig, Igf2prox and ICR anchors). These looping interactions (S1 Fig, boxed region; Sheets A-D in S5 Table) bring the skeletal muscle enhancer (SME) [5] into proximity of Igf2 on the paternal allele and H19 on the maternal allele, facilitating allele-specific expression of each promoter in skeletal muscle. Next, we confirmed biallelic looping between CCD and Igf2prox in WT skeletal muscle, consistent with what was reported from syntenic regions in human cells (S1 Fig, purple bars, CCD anchor [non-allelic], Igf2prox anchor [bi-allelic]) [30]. To understand how CCD-dependent architecture may repress Igf2 expression in adult skeletal muscle, we next examined the architectural impact of the CCDIvr deletion.

Elevated and altered looping interactions on ΔCCDIvr alleles in adult skeletal muscle

We investigated whether, in the absence of CCD, new looping interactions emerge from the Igf2prox-anchored region. Igf2prox anchored loops brought the SME and Igf2 promoter into proximity on the paternal allele (S1 Fig). We hypothesized that this anchor could capture new enhancer-promoter (E-P) interactions on ΔCCDIvr alleles, which would explain elevated Igf2 expression in adult skeletal muscle. First, CCD anchored looping to Igf2prox was observed from the WT alleles in both maternal and paternal CCDIvr deletion nuclei, indicating biallelic CCD-Igf2prox looping (Fig 3E, CCD anchor). As expected, on both maternal and paternal ΔCCDIvr alleles, Igf2prox looping to the deleted CCDIvr region was absent. However, no new Igf2prox-anchored looping interactions were detected (S3C Fig, neonatal; Fig 3E, adult [boxed region at CCD]; Sheet E in S5 Table). Finally, in adult skeletal muscle, Igf2prox-anchored interaction peaks to CTCF sites near SME were higher on the paternal ΔCCDIvr allele compared to the corresponding paternal WT allele (Fig 3E, boxed region surrounding SME; Sheet E in S5 Table). This may occur because loss of the competing CCD/Igf2prox loop increases chances for SME/Igf2prox looping. The increased proximity of Igf2 to SME (S3E Fig) likely contributes to elevated Igf2 expression on the paternal ΔCCDIvr allele in adult skeletal muscle.

To directly measure Igf2 E-P interactions, we next evaluated interactions captured from an Igf2 promoter anchor probe (12 kb centromeric to the Igf2prox anchor [Sheet A in S4 Table]). This region was previously shown to loop to the SME on the paternal allele using chromatin conformation capture (3C) in neonatal skeletal muscle [31]. Consistently, the Igf2 promoter anchor (blue bar) captured peaks at the SME (green bar) on the paternal but not the maternal WT alleles in skeletal muscle (S4A Fig [neonatal]; Fig 4A [adult]; Sheets F and G in S5 Table). Igf2 promoter/SME looping was also captured on the paternal ΔCCDIvr alleles (Fig 4A; S4A Fig). In adult muscle, SME interaction peaks were higher on the paternal ΔCCDIvr versus WT allele (Fig 4A, boxed region; Sheet F in S5 Table), demonstrating that deleting CCD increases the probability of SME to engage Igf2 promoters. Interestingly, a new Igf2 promoter anchored-interaction to a sequence between ICR and CCD was detected on the paternal ΔCCDIvr allele (Fig 4A, boxed region; Sheet F in S5 Table). A stronger paternal-specific Igf2 promoter looping interaction to this sequence may explain the observed strong increase in Igf2 expression on the paternal versus the maternal ΔCCDIvr allele in adult skeletal muscle. Thus, we hypothesized that this sequence between ICR and CCD functions as a developmental- and tissue-specific enhancer to drive high Igf2 expression on both the maternal and paternal ΔCCDIvr alleles.

Fig 4. Evidence for a Igf2 PaSME between the ICR and CCD in adult skeletal muscle.

(A) Normalized allele-specific Capture-C peaks on WT, MKO and PKO alleles. Captured interactions from anchor at Igf2 promoter (blue bar) reveals enhancer-promoter (E-P) interactions on the paternal PKO allele (Sheet F in S5 Table). (A) Boxed regions at SME and sequence between ICR and CCD (?) are significant E-P interactions present on the paternal PKO allele as measured by Wald statistics using two methods (* or # p < 0.05, ** p < 0.01, Sheet F in S5 Table). (B) Sequence between ICR and CCD with hallmarks of an adult SkM enhancer (PaSME) (Sheet A in S2 Table). (1) ENCODE DHS data show an adult SkM specific peak at PaSME (green bar). (2) Enhancer mark H3K27ac, H3K4me2 and ATAC peaks at PaSME in adult SkM. (3) ENCODE MYOD1 and MYOG ChIP-seq peak in differentiated myocytes coincide with PaSME. (C) Allele-specific H19 expression from maternal and paternal WT versus ΔCCDIvr alleles in adult GAS. Allele-specific H19 (D) and Igf2 (E) expression from maternal and paternal WT versus ΔICR alleles in adult GAS. Paternal-specific H19 levels from WT versus paternal ΔICR allele (D) and maternal-specific Igf2 levels from WT versus maternal ΔICR allele (E). For (D) and (E), two-tailed Welch’s t-test: **p < 0.01; error bars represent SD.

Evidence for a putative adult skeletal muscle enhancer (PaSME) between the ICR and CCD

To identify adult skeletal muscle enhancers at the Igf2/H19 locus, we analyzed publicly available data characteristic of open chromatin, including DHS, assay for transposase-accessible chromatin with sequencing (ATAC) data, and histone enhancer marks (ChIP-seq H3K27ac and H3K4me2) across multiple tissues (Sheet A in S2 Table). Mouse skeletal muscle enhancer regions have been shown to be associated with coinciding H3K4me2 and H3K27Ac marks [38,39].

In neonatal/embryonic tissues, DHS marks the known SME in embryonic limb and neonatal heart, but not in neonatal liver (S4B Fig, green bar at SME). Similarly in embryonic limb, the putative enhancer region (~2 kb) captured by the Igf2 promoter anchor on the paternal ΔCCDIvr allele in adult skeletal muscle (Fig 4A) demonstrated a modest limb-specific DHS peak that coincides with H3K27ac enrichment (S4B Fig). In adult skeletal muscle, this region exhibits stronger H3K27ac, H3K4me2, DHS and ATAC peaks, suggesting preferential enhancer activity in adults [38,40]. Furthermore, this sequence colocalizes with muscle-specific transcription factors and known markers of muscle-specific enhancers MYOD1 and MYOG during differentiation of C2C12 skeletal myoblasts to myotubes (Fig 4B and Sheet A in S2 Table). Taken together, these data suggest there is a ~ 2kb putative adult skeletal muscle enhancer sequence (designated PaSME) between the ICR and CCD (Fig 4B; S4B Fig, green bar at PaMSE). Collectively with Capture-C data, we hypothesize that 1) on the WT alleles, the CCD/Igf2Prox interaction blocks Igf2 from accessing PaSME and 2) on both the maternal and paternal ΔCCDIvr alleles, PaSME is responsible for Igf2 activation.

Mouse 2D and 3D architecture at the TAD encompassing Igf2/H19 locus is conserved in human cells

We next ascertained whether there was evidence for a PaSME-like region in human cells using publicly available human 3D and 2D chromatin data at the IGF2/H19 locus. We retrieved 3D architecture data {chromatin interaction analysis by paired-end tag sequencing (ChIA-PET) [41]} visualized with Juicebox.js software (Sheet B in S2 Table), which depicts a CCD-CTCF dependent IGF2 proximal-CTCF interaction (S5A Fig, dashed boxes; Sheet B in S2 Table) [42]. Secondary analysis further showed that CTCF dependent looping interactions from the H19 ICR CTCF sites are maternal and interactions from the IGF2 proximal CTCF sites are paternal [41]. CTCF site orientations and ChIP-seq binding are conserved at mouse Igf2/H19 locus (S1 versus S5A Fig; Sheets D versus E in S2 Table). At the human ICR, one of CTCF sites originally defined [43] is not functional as defined by ChIP-seq binding (S5B Fig and Sheet E in S2 Table).

We also used publicly available DHS data across multiple adult tissues and ChIP-seq data for enhancer marks (H3K27ac, H3K4me2, H3K4me1) to identify muscle specific enhancers. Remarkably, adult muscle-specific enhancer marks and open chromatin at the murine PaSME are conserved at the orthologous human region in adult skeletal muscle (S5A Fig and Sheets B and C in S2 Table). Publicly available 2D data also indicate there is a putative adult skeletal muscle specific enhancer (PaSME) between the ICR and CCD at the human IGF2/H19 locus (S5A Fig). Nevertheless, in both human and mouse the function of PaSME when CCD is intact, remains to be determined.

Access to PaSME is not associated with high activation of nearby genes H19, Nctc1 and Mrpl23 in adult skeletal muscle

We hypothesized that (on the maternal allele) the ICR blocks the PaSME from engaging the H19 promoter, analogous to how CCD blocks PaSME from engaging Igf2 promoters. As expected, H19 expression was unchanged on ΔCCDIvr alleles in neonatal and adult muscle (Fig 4C; S2D, S2E Fig). However, we anticipated H19 would be highly elevated on ΔICR alleles in adult skeletal muscle, if unblocked PaSME engaged the H19 promoter.

In neonatal muscle, H19 and Igf2 are expressed from the paternal and maternal ΔICR alleles at levels less than or equal to the WT alleles through sharing of the SME [5,6,26,27,31]. In adult GAS, H19 and Igf2 were modestly activated/elevated from the paternal and maternal ΔICR alleles, respectively (Fig 4D, 4E). (H19 and Igf2 levels from the respective maternal and paternal ΔICR alleles did not appear elevated.) These results suggest H19 and Igf2 on the ΔICR alleles are not under the control of an adult skeletal muscle specific enhancer activity, either from SME or PaSME. The outcomes indicate the PaSME (1) engages specific promoters (Igf2 but not H19), or (2) activity is unidirectional, only engaging telomeric promoters.

We anticipated that genes immediately proximal to H19 (non-coding Nctc1 and Mrpl23) may be more accessible to the WT paternal PaSME compared to maternal PaSME due to the CTCF-bound maternal ICR functioning as an enhancer-blocker. Consistent with this expectation, PaSME anchored interactions to the Igf2 promoter region were enriched on the adult paternal ΔCCDIvr allele (S4B Fig). Indeed, our data shows enrichment of interactions between paternal PaSME and the region surrounding Nctc1 on the WT paternal allele compared to the maternal allele. If paternal PaSME were engaging with the Nctc1 or Mrpl23 promoters in a similar manner to the established enhancers for Igf2/H19, then we hypothesized that Nctc1 and Mrpl23 expression would be skewed towards the paternal allele. However, Nctc1 and Mrpl23 were biallelically expressed in adult skeletal muscle [31,44,45], suggesting that there are different regulatory elements contributing to Nctc1 and Mrpl23 expression.

Discussion

Here we uncover a mechanism by which CCD may repress Igf2 by insulating Igf2 from a PaSME. We first characterized the lncRNA, Ivr, adjacent to CCD, and determined that Ivr’s preferential paternal allele expression is likely due to sharing of Igf2 enhancers that are centromeric to H19. Further, Ivr alone was not required to maintain CCD’s skeletal muscle Igf2 repressor function. Deleting both the CCD and Ivr (ΔCCDIvr allele) partially activated maternal Igf2 in neonatal skeletal muscle but generated extraordinary up-regulation of both maternal and paternal Igf2 in adult skeletal muscle. Temporal and tissue specific 2D chromatin architecture data were used to identify a PaSME between ICR and CCD that is conserved at the syntenic human region. This 2D analysis, together with our allele-specific 3D architecture analysis of WT and ΔCCDIvr alleles, led us to propose that CCD dependent architecture blocks Igf2 access to the PaSME located between the ICR and CCD. Below, we discuss properties of CCD (versus Ivr) and our model invoking cohesin loop extrusion and enhancer synergy to explain how Igf2 accessibility to skeletal muscle enhancers SME and PaSME directs Igf2 expression from WT and ΔCCDIvr alleles (Fig 5).

Fig 5. Model of Igf2 accessing skeletal muscle enhancers (SME and PaSME) in adult SkM on maternal and paternal WT versus ΔCCDIvr alleles by cohesin loop extrusion and enhancer synergy.

For cohesin complex loaded at regions 1, 2 and 3 (above alleles), cohesin loop extrusion model predicts formation of specific loops on the maternal (pink) and paternal (blue) alleles. Maternal ICR looping demarcates maternal H19 exclusive access to the SME on both the WT and ΔCCDIvr maternal alleles, whereas CCD and Igf2prox looping insulates Igf2 access to PaSKM on WT allele. We propose that an Igf2 promoter gains access to PaSME (dashed line with arrow) on the maternal ΔCCDIvr allele. WT paternal Igf2prox forms loop with CCD (loop 3) and/ or with Mrpl23 CTCF sites (loop 1/2). SME most likely engages Igf2 SME within loop ½. On the paternal ΔCCDIvr allele, there is (1) increased probability of looping between Igf2prox and sequence near SME and (2) there is increased probability for an Igf2 promoter to engage with SME and PaSME.

With the number of lncRNAs surpassing protein coding genes, efforts have focused on the role of lncRNAs [21]. Due to their low sequence conservation, elucidating lncRNA function is challenging [46]. As such, we did not find evidence for IVR-like transcription or gDNA conservation in the orthologous region adjacent to the conserved human CCD sequence. Whereas gene regulatory roles for several lncRNAs in cis and trans have been uncovered [21,46–50], Ivr was not required for normal allele-specific Igf2 expression in neonatal and adult skeletal muscle (Fig 2A; S3B Fig, ΔIvr allele). For non-functional lncRNA transcripts, the regulatory region required for transcription, not the transcript, may dictate function [21,48,51]. Thus, we propose that cohesin and CTCF-bound Ivr promoter/regulatory region, CCD, may regulate Igf2 expression in skeletal muscle as an architectural insulator (Fig 5 model). We cannot discount that the deleted Ivr sequence contributes to the ΔCCDIvr allele phenotypes, although deleting CCD sequence alone activated maternal Igf2 in an early transgenic model [17].

Loop extrusion provides an explanation for the formation of cohesin guided chromatin CTCF loops, whereby cohesin complexes are loaded at potentially random sequences, and then loops of DNA are extruded until they are connected/stalled at convergent CTCF sites/boundary elements [35,52,53]. At the Igf2/H19 locus, in neonatal and adult skeletal muscle, WT maternal and paternal-specific loops anchored from the ICR and Igf2prox CTCF sites (S1 Fig) were as predicted (Fig 5; maternal WT allele CTCF loops 1 and 3, paternal WT allele CTCF loops 1/2 and 3). On the maternal WT alleles, CTCF bound ICR forms a strong block to cohesin migration as loop interactions with Igf2prox anchor appear limited to CCD on the maternal WT allele (S1 Fig) and absent on the maternal ΔCCDIvr allele (Fig 3E; S3C Fig).

On the paternal allele, CTCF does not bind the hypermethylated ICR [54] and Igf2prox/Mrpl23 loops are formed (Fig 5 -paternal allele 1/2 loop, S1 Fig).On both parental alleles, cohesin loaded at sequence between CCD and Igf2 is predicted to direct the CCD/Igf2prox looping (Fig 5 [Loop 3]; S1 Fig). This Igf2prox/CCD CTCF loop insulates Igf2 within the loop from enhancers outside of the loop. Because there are multiple CTCF sites at Igf2prox, the Igf2prox/CCD loop could form within the Igf2prox/Mrpl23 loop (Fig 5, WT paternal alleles). Our Capture-C analysis cannot determine if these two Igf2prox anchored loops were formed simultaneously in one cell or individually in different cells.

It is agreed that (1) E-P interactions require physical proximity and (2) enhancers and target genes should be within the same CTCF loop domain [55,56]. The SME engages maternal H19 within the ICR anchored loops and paternal Igf2 within Igf2prox anchored loops (Fig 5; S1 Fig). In contrast, the hypothesized PaSME, external to the Igf2prox/CCD loop, is blocked from engaging Igf2 on both parental alleles. We propose that activation of Igf2 on the maternal ΔCCDIvr allele in adult skeletal muscle is due to removal of the block to adult skeletal muscle enhancer activity (PaSME) (Fig 5). On the maternal ΔCCDIvr allele (Fig 3E; S3C Fig), our data suggest that no new Igf2prox anchored loops are formed that would facilitate PaSME engaging Igf2 promoters. On the paternal ΔCCDIvr (versus WT) allele in adult skeletal muscle, Igf2prox/Mrpl23 interactions and Igf2 promoter contacts to both SME and PaSME are strengthened (Fig 3E, 4A, 5). We propose that high Igf2 activity on the paternal versus maternal ΔCCDIvr allele (Fig 3D versus 3B) is due to SME and PaSME acting in synergy to activate adult Igf2 (Fig 5) [55,57].

Because CTCF loops are dynamic and some E-P interactions are not cohesin-dependent, loop extrusion does not fully explain E-P interactions [52,58,59]. (1) How might CCD block PaSME activity independent of Igf2prox/CCD loop formation on the paternal WT allele (Fig 5, Loop 1/2)? (2) How does PaSME access Igf2 promoters on the ΔCCDIvr allele? The Igf2 promoter/PaSME interaction could be mediated by CTCF/cohesin-independent mechanisms that bridge E-P interactions by use of self-interacting DNA associated proteins and transcription factor hubs at enhancer regions [60]. Furthermore, deconstructing the regulatory controls of enhancer-bound and promoter-bound proteins that cause these elements to interact may uncover why access to PaSME results in high levels of Igf2 activation on the ΔCCDIvr alleles (Fig 3B, 3C, 3D), but minimal change in H19 expression on the ΔICR alleles (Fig 4C) [56]. Alternatively, PaSME activity may be orientation-dependent, as has been demonstrated for an enhancer at globin locus [61], and therefore unblocked PaSME engages with telomeric Igf2 but not centromeric H19. We anticipate advancement of high-resolution live cell imaging technologies that simultaneously report between 3-D architecture and gene transcription in cells of interest are required to decipher these predicted outcomes [59,62]. Deletion of PaSME in the presence and absence of CCDIvr will validate the role of PaSME in adult skeletal muscle.

The ΔCCDIvr mice are a model for how tissue-specific overexpression of Igf2 can affect skeletal muscle development in adults, when Igf2 is normally repressed. In vitro studies have shown that Igf2 plays a role in myoblast differentiation, however, our understanding of Igf2 in muscle development and regeneration in vivo is limited [63,64]. Global prenatal transgene Igf2 overexpression led to perinatal lethality, whole body overgrowth and other Beckwith-Wiedemann syndrome phenotypic determinants [65]. Transgenic models with persistent adult tissue-specific Igf2 overexpression led to localized overgrowth and/or cancer, uncovering autocrine and endocrine mechanisms of action [66,67]. In a Duchenne Muscular dystrophy mouse model, muscle degenerative phenotypes associated with lack of dystrophin were ameliorated by elevated serum IGF2 peptide from one of these transgenes [67,68]. While some studies have suggested that approaches to elevate muscle IGF2 peptide availability could be therapeutic for the treatment of muscular dystrophies [64,68], one indicated it would be detrimental [9]. Defining metabolic and physiological outcomes and IGF2 autocrine signaling consequences of Igf2 overexpression in ΔCCDIvr adult skeletal muscle could resolve these possibilities.

In summary, we demonstrate that CCDIvr sequence functions as a skeletal muscle repressor by forming a loop with Igf2 proximal CTCF sites and thereby isolating Igf2 within the loop from skeletal muscle enhancers outside of the loop. The Igf2/H19 locus continues to be an excellent model for deciphering how allele-specific genome architecture facilitates enhancer-promoter interaction that dictate tissue-specific gene expression.

Limitation of Capture C analysis: (1) Few deletion allele samples were analyzed, therefore we could not evaluate statistical significance for the neonatal maternal and paternal ΔCDDIvr alleles and the adult maternal ΔCDDIvr allele. (2) From the capture-C libraries we achieved a read length up to 300nt (we had anticipated reads up to 500nt). The achieved read lengths were not long enough to capture connections with Mrpl23 and SME CTCF site anchored probes (Sheets A and E in S4 Table) Nevertheless, from the Igf2prox, Igf2 promoter, CCD and PaSME anchored probes we observed similar changes interactions peaks on individual ΔCDDIvr alleles (compared to the WT alleles) that were measured for the adult paternal ΔCDDIvr (2 samples, S5 Table).

Materials and methods

Ethics statement

Mouse studies were approved by and performed in accordance with the Institutional Animal Care and Use Committee (IACUC) at the University of Pennsylvania (Protocol Number: 804211).

Animal

Briefly, male and female C57BL/6J (JAX:00664) (B), E2a-Cre (JAX:003724) and ROSA26::FLPe (JAX:016226) mice were obtained from Jackson Laboratories to generate new alleles (ΔIvr + Neo, ΔIvr and ΔCCDIvr). Cast7 (C) mice were also sourced from Jackson Laboratory and used to generate progeny for allele-specific analysis [69]. Mouse model Δ3.8kb-5’H19 (here designated ΔICR) [26] was used for comparative analyses with our newly derived ΔCCDIvr line. All mice with the deletion alleles (mouse lines ΔIvr + Neo, ΔIvr and ΔCCDIvr) were maintained on a B background and characterized as described below. These mice were backcrossed into C57BL/6J for at least 6 generations. When not breeding, up to five adult mice were housed in individually ventilated cages with a 12-hour light/dark cycle at a temperature of 22°C and relative humidity of 50%. All mice were maintained in a pathogen-free facility and housed in polysulfone cages. Mice had ad libitum access to standard laboratory chow (Laboratory Autoclavable Rodent Diet 2010, LabDiet), drinking water, and environmental enrichment (i.e., nesting material was provided to enhance their welfare).

Beginning at weaning, body weight measurements of adult mice were recorded, on average, every 2 weeks. Prior to tissue collection, neonates were euthanized by decapitation and adults were euthanized by cervical dislocation. For initial survey analysis, skeletal muscle subtypes including gastrocnemius, quadriceps, tibialis anterior and tongue were collected from 7 weeks-old male mice (Fig 2D). For subsequent analyses, the skeletal muscle subtypes extensor digitorum, gastrocnemius, quadriceps, soleus, tibialis anterior and tongue were collected from 4-5 month-old adult mice. For expression analyses, a minimum of 3 mice of each genotype (WT and maternal- or paternal-specific inheritance of deletion alleles) were used. Adult male mice tissues were processed for expression analyses. All tissues were snap-frozen when collected, unless used for nuclei isolation as described below. Frozen tissues were stored at -80°C.

Target vector

The target vector was designed to delete the Ivr and introduce loxP sites for Cre-mediated deletion of CCD. The target vector arms (5’ arm - included CCD sequence [mm10 chr7: 142,604,115–142,611,247]; 3’ arm – included sequence adjacent to Ivr coding sequence [mm10 chr7:142,620,880–142,624,437]) were amplified from 129SvEvBAC genomic DNA (bMQ318o12) by iProof High Fidelity DNA Polymerase (BIO-RAD, Hercules CA USA) and cloned into BluescriptIIKS. Following published guidelines [70–72], recombineering was used to introduce (1) a loxP site on the target vector 5’ arm (replaced G at mm10 Chr7: 142608373) and (2) a Frt-flanked neo cassette plus a loxP site between the 5’ and 3’ target vector arms. Finally, for negative selection, a 2.3kb SalI diphtheria toxin A cassette (DTA) [73] was cloned adjacent to the 3’ target vector arm. The vector was linearized by NotI digestion. See S2 Fig for the target vector schematic.

TALEN plasmids

We assembled transcription activation-like effector nuclease (TALEN) plasmids to facilitate targeting at the Ivr locus. We used Paired Target Finder to design TALENs to minimize off-target recognition [74] (https://tale-nt.cac.cornell.edu/). TALENs were designed to target double-stranded breaks of endogenous sequence that were not present in the target vector (i.e., 5’ target vector arm where loxP site is inserted and Ivr sequence that is designed to be deleted by the targeting event). We assembled TALENs using the TALE Toolbox system [75]. (See S2 Fig for location of TALEN targets: on 5’arm TALENs were generated for targeting ‘caccccctttccttgctga’ and ‘tttctttcatggccccta’; on 3’ arm TALENs were generated for targeting ‘caccccctttccttgctga’ and ‘tttctttcatggccccta’.)

Generation of ΔIvr + neo allele in ESCs

Linearized target vector was electroporated into E14.1 ESCs [76] to generate ΔIvr + neo ESCs using previously described cell culture, electroporation and Southern blot analysis methods [77], with the following exceptions. To facilitate targeting, TALEN plasmids targeting CCDIvr locus were electroporated with the target vector (S2 Fig). Genomic DNA (gDNA) isolated from G418-resistant positive clones was digested with DraI and analyzed by Southern blots with 5’ and 3’ external probes (S2 Fig and Sheet A in S3 Table).

Generation of mice with the ΔIvr + Neo, ΔIvr and ΔCCDIvr alleles

ESC clones heterozygous for the ΔIvr + Neo allele were provided to the Transgenic and Chimeric Mouse Facility in the Department of Genetics at the University of Pennsylvania for injection into blastocysts to derive chimeric mice. Chimeric mice were bred to C57BL/6J and agouti progeny inheriting the ΔIvr + Neo allele were bred to ROSA26::FLPe knock-in mice (RRID:IMSR_JAX:016226) to remove the Frt-flanked neomycin-resistance cassette (neo) to generate progeny with the ΔIvr allele. Mice with the ΔIvr + Neo allele or ΔIvr allele were bred to E2a-Cre transgenic mice (RRID:IMSR_JAX:003724) to remove loxP-flanked CCD sequence to generated progeny with the ΔCCDIvr allele.

DNA isolation and genotyping

DNA was extracted from ESCs [77], mouse ear punches [78] and mouse tissues [79]. Primers and conditions used for PCR genotyping E2a-Cre, ROSA26::FLPe, ΔICR, ΔIvr + Neo, ΔIvr and ΔCCDIvr alleles are provided in Sheet B in S3 Table [26,80]. The maternal allele is listed first and the paternal allele is listed second for all genotypes.

RNA isolation

RNA was extracted from frozen tissues in TRIzol (15596026, Thermo Fisher Scientific, Waltham, MA, USA) using manufacturer’s instructions. This included homogenization of 50–100 mg tissue in 1 ml TRIzol with a Polytron homogenizer (Binkmann Polytron Kinematica PT 10–35) for 25–30 seconds, optional centrifugation of homogenate to collect cleared supernatant and optional addition of 1 μl GlycoBlue (AM9516 Invitrogen) to the aqueous phase following chloroform extraction. RNA was then precipitated and purified from the aqueous phase on a RNAeasy Mini kit column following manufacturer’s instructions (Qiagen, Hilden, Germany). gDNA was removed from RNA using RQ1 RNase-Free DNase (Promega, Madison, WI, USA) using manufacturer’s instructions.

Ivr 5’RACE-PCR

In the laboratory of Dr. Denise Barlow [22], 5’RACE was performed with FirstChoice RLM-RACE Kit, with two rounds of nested PCR for 5’RACE. 5’RACE-PCR products were gel extracted and then cloned into pGEM-T for sequence analysis. See Sheet A in S1 Table for 5’RACE primers and pGEM-T 5’RACE clone sequences.

cDNA synthesis, allele-specific expression and total RNA

cDNA synthesis, reverse transcription quantitative polymerase chain reaction RT-qPCR and allele-specific expression analyses were performed as previously described [81]. Allele-specific expression primers with RFLP outcomes are provided in Sheet C in S3 Table [81,82]. Total RNA levels for all genes were quantified by RT-qPCR and measured relative to the geometric mean of RNA levels of Arbp (acidic ribosomal phosphoprotein P0), Nono (non-POU domain-containing, octamer-binding protein) and Rpl13a (ribosomal protein L13A). For RNA levels reported on y-axis, the mean value of wild-type measurements was set as 1. Primers for RT-qPCR are provided in Sheet D in S3 Table [81].

Statistical analysis or analyses

Differences between two groups were evaluated by unpaired t test with Welch’s correction using GraphPad Prism software: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. For Capture C data, statistical analyses are described below.

CTCF binding site and orientation analysis

The most likely CTCF binding sites within DNA fragments (~150–300 bp) within predominant CTCF ChIP-seq peaks (across mouse region mm10 Chr7:142,300-000-142,750,000 and at human region hg19 Chr11:1,645,000–2,225,000) were analyzed. The most likely CTCF binding and CTCF motif orientation was determined using the PWMScan – Genome Position Weight Matrix (PWM) scanner (https://epd.expasy.org/pwmtools/pwmtools/pwmscan.php) [83] and CTCF consensus sequence from the JASPAR CORE 2020 motif library [84] to extract the matching score for the best motif instance at each binding site. Human CTCF motif orientation was also determined with the PWMScan. See Sheets D and E in S2 Table for method details.

Nuclei isolation for capture C

Nuclei were isolated from fresh neonatal and adult mouse hindlimb muscle from hybrid WT mice (BxC and CxB) and ΔCCDIvr mice (ΔCCDIvr x C and C x ΔCCDIvr) (Sheet B in S4 Table). Neonatal hindlimb nuclei were isolated as described by Juan et al. 2022 with a few modifications [85]: Isolated hindlimb muscle was placed on ice in a 1.5ml Eppendorf tube, covered with ice-cold homogenization buffer (10 mM Tris-HCl, pH 8.0; 5 mM CaCl₂; 3 mM MgAc₂; 0.32 M sucrose; 0.1% Triton-X; 0.1 mM EDTA, 1X Protease Inhibitor cocktail without EDTA (Roche); 1mM DTT; 0.1 mM PMSF) and then minced using a fine scissor. Minced muscle was transferred to 10 ml homogenization buffer in a 15 ml tube and dounced 21 times with loose-fitting pestle and then 14 times with tight -fitting pestle. Released nuclei were cleared of debris by filtration into 50 ml conical tubes through a series of pre-rinsed cell strainers (100 μm, 70 μm, 40 μm and 20 μm) as adapted from Nohara et al. 2017 [86]. At room temperature, the cleared nuclei were then fixed on a rotating platform in 1% formaldehyde for 10 minutes and fixation was stopped with glycine at a final concentration of 0.125 M on rotating platform for 5–10 minutes. Nuclei were pelleted at 500 x g for 15 minutes at 4°C. Then nuclei were washed with 50 ml ice-cold PBS and again centrifuged at 500 x g for 15 minutes at 4°C. Nuclei were resuspended in ~2 ml cold PBS, counted, transferred to a LOW BIND Eppendorf (≥ 5 ml) tube and centrifuged at 700 x g for 10 minutes at 4°C. Pellet was snap frozen in liquid nitrogen and stored at -80°C.

All steps to isolate adult hindlimb nuclei were performed in a cold room with reagents stored and procedures performed on ice, except for fixing of nuclei. Gastrocnemius, tibialis anterior and quadriceps muscle were dissected, placed in 2ml tube, covered with freshly prepared hypotonic buffer (10 mM HEPES-KOH (pH 7.3, 10 mM KCl, 5 mM MgCl₂, 0.1% NP-40, 0.1 mM PMSF, Protease Inhibitor cocktail (PIC) 1X, 1 mM DTT) and then minced with fine scissors. The methods and hypotonic buffer (with addition of 1 mM DTT) of Joshi et al. 2017 [87] were used, with slight modifications. Minced muscle was added to a 50 conical tube with 10 ml hypotonic buffer and homogenized with a mechanical tissue homogenizer for ~10 seconds. Hypotonic buffer was added to dilute homogenate to 30 ml, homogenate was added to 40 ml dounce tube and incubated for 5 minutes on ice and then dounced with loose pestle 10 times. Lysate was divided into two 50 ml conical tubes, transferred to room temperature, and fixed as described above for the neonatal nuclei isolation. The tubes of fixed lysate were recombined and then returned to the cold room where it was further dounced 10 times with the loose pestle, pelleted at 1,000 x g for 5 minutes at 4°C and then resuspended in 10 ml of hypotonic buffer. Lysate was then filtered through cell strainers as described above for neonatal nuclei, adapting adult muscle nuclei filtration protocols [86,87]. Nuclei were counted and filtrate was transferred to a 15 ml LOW BIND Eppendorf tube, centrifuged at 1,000 x g for 5 minutes at 4°C. Pellet was snap frozen in liquid nitrogen and stored at -80°C.

Next generation Capture-C sequencing and statistical analysis Capture-C The Capture-C was as described [85]. Nuclei were isolated from neonatal and adult hindlimb muscle obtained from reciprocal hybrid WT mice (BxC and CxB) and heterozygous mutant ΔCCDIvr mice. We constructed, fragmented and indexed 3C library, performed oligonucleotide capture on pooled libraries, and performed sequencing and analyses of captured sequences, as previously described [85], with the following modifications. Here, the biotinylated oligonucleotides were designed for multiple CTCF binding sites across the Igf2/H19 locus (including the ICR), Igf2 promoters and previously identified and putative enhancer skeletal muscle enhancers using IDT Discovery Pool technology (Sheets A and C in S4 Table). The first hybridization reaction included five pools described by Juan et al. and two additional pools described herewith (one pool assayed the four neonatal muscle samples; the other pool assayed the six adult muscle samples [see Sheets D and E in S4 Table]), which were each processed as previously described [85].

Bioinformatic analysis of the resulting reads was conducted as previously described [85], with the following modifications to implementing the CCseqBasicS pipeline. The CCseqBasic tool (vCC5) tool was used in both non-SNP-specific as well as SNP-specific configurations, however, here C57BL/6J (B)/Castaneous (C) SNPs were evaluated. (Most designed probes had B/C SNPs nearby to be sequenced; the list of used SNPs is available in Sheet C in S4 Table). The total read counts for each probe region, including non-SNP-specific regions and regions for each SNP (C, B), are provided in Sheet D in S4 Table. Bigwig files were generated for each sample and each probe by piling up all reads, B reads and C reads separately, and then normalized to 10K reporter reads per designed probe for track visualization. We pooled reads from alleles in the same experimental group (WT.neonatal.maternal, WT.neonatal.paternal, WT.adult.maternal, WT.adult.paternal, [CCDIvr pKO] KO.adult.paternal) together for calling contact sites, statistical analysis and visualization (Sheets K and L in S5 Table). For each experimental group and for each probe, we defined contact sites in two ways: the “given coordinates” approach which we manually defined contact sites (loci with CTCF motif, Igf2 promoters, previously reported skeletal muscle enhancer and a putative skeletal muscle enhancer [Sheets A-J in S5 Table]), and the “peak calling” approach for which we used computational algorithm to call contact sites from the data. More specifically for the peak calling approach, contact sites were defined as peaks called using MACS2 (v2.1.1) with parameters “--nomodel --extsize 1200 --qvalue 1e-100 --broad --nolambda --broad-cutoff 1e-100 --minsize 5000”. The corresponding contact site visualization, quantification of contact sites, differential analysis between experimental groups and generation of figures on WashU Epigenome browser was performed as previously described [85]. Statistics for locus-specific comparisons are provided in S5 Table for both the “peak calling” and “given coordinates” approach.

Data and code availability

The accession number for raw and processed Capture-C sequencing data reported in this paper is publicly available on date of publication and is deposited at GEO:GSE202669. We examined existing, publicly available data provided by the ENCODE Project using the New WashU Epigenome Browser, visualized on the mouse mm10 and human hg19 genomes (Sheets A and B in S2 Table). Additional supporting data is provided in S1 and S2 Data files. We do not report original code.

Supporting information

S1 Fig. CCD association with Igf2 within the Igf2/H19 locus TAD (UCSC browser mm10 chr7:142,350,000–142,700,000).

From top to bottom, the following are depicted: Genes, regulatory elements (SME, EE, ICR and CCD) and probes are annotated (Sheet C in S2 Table). Ivr overlaps with predicted (GM33148) lncRNA. CTCF binding site (BS) polarity (green forward arrow/bars for forward, and red reverse arrow/bars for reverse) is designated with WashU Epigenome Browser tracks using Public ENCODE and 4D Nucleome Network data (Sheet A in S2 Table). Micro-C interactions between CCD and CTCF sites proximal to Igf2 (Igf2prox) are highlighted (dashed lines). Purple arrow heads mark Igf2prox interaction that are detected by the following Capture-C analysis. Below are normalized Capture-C interaction frequencies of WT F1 hybrid neonatal and adult skeletal muscle from the viewpoints probes (anchors) at CCD, at Igf2prox, and at the ICR (Sheets A-D in S5 Table). Interactions between CCD (biallelic) and Igf2prox (maternal and paternal) are highlighted (purple vertical bars). Black tracks designate interactions from anchor (CCD) without an available polymorphism to discriminate alleles. Pink and blue tracks designate maternal and paternal alleles, respectively, and darker intensities correspond to stronger interactions. Selected allele-specific interactions are boxed (paternal Igf2prox to the SME region, maternal ICR to the SME region, maternal proxKrtap5–4 to ICR and paternal proxKrtap5–4 to proxIgf2) with Wald statistics using two methods (* or # p < 0.05, ** or ## p < 0.01, *** or ### p < .001, #### p < .0001, Sheets A-D in S5 Table). Probe and regulatory regions are also annotated below the Capture C data.

(PDF)

S2 Fig. Generation of ΔIvr and ΔCCDIvr alleles and neonatal analysis.

(A) UCSC genome browser view of Igf2/H19 locus with mapped CCD, conserved regions CR1 and CR2, DHS, IvrEx1, IvrEx2 and IvrEx3 sequence (Sheet C in S2 Table), above, and targeting vector and targeted loci ΔIvr + neo, ΔIvr (after Flp mediated removal of neo) and ΔCCDIvr (after Cre mediated removal of CCD), below. Rr1^em1.1Msb, Rr1^em1.2Msb and Rr1^em1.2Msb are the MGI approved nomenclature for the respective targeted alleles. The endogenous locus (thick black line) depicts endogenous DraI sites, the sites where DraI and loxP sequences (pink bars) were introduced (insert site), TALEN cut sites (T), CCD (purple box), Ivr transcription unit (gray boxes), targeting vector 5’ and 3’arms (open boxes) and the 5’ and 3’ sequence used for Southern blot analyses (5’P and 3’P, dark gray boxes). In addition to the 5’ and 3’ arm (thick black line), the targeting vector includes introduced DraI site, loxP sites (pink bars), Frt (blue bars) flanked Neo-cassette (open box), DT cassette (open box), and BSSKII (dashed line). Primer pairs (c, d, e, f) used to genotype the targeted loci before and after neo and CCD removal are depicted below the endogenous and target loci. (B) DraI digested DNA analyzed by Southern blots. Using the 5’P probe, the 18.2kb endogenous and 8.9kb targeted DraI fragments are detected. Using the 3’P probe, the 7.2 endogenous and 8.9kb, 7.1kb and 4.1kb target DraI fragments are detected. (C) Total Igf2 expression in CCDIvr^+/+ versus CCDIvr^+/Δ (paternal ΔCCDIvr) neonatal tongue. (D) Allele-specific H19 expression in neonatal liver (Li), tongue (T) and heart (H) from WT and ΔCCDIvr alleles. H19 qRT-PCR products are digested with Cac8I to discriminate the parental B (WT B, ΔCCDIvr) and C alleles. (E) Total H19 expression in CCDIvr^+/+ versus CCDIvr^Δ/+ (maternal ΔCCDIvr) neonatal tongue. (F) Weights of WT (CCDIvr^+/+) versus maternal deletion (CCDIvr^Δ/+) and WT (CCDIvr^+/+) and paternal deletion (CCDIvr^+/Δ) neonatal littermates. No statistical difference between weights of mutant and WT littermates.

(PDF)

S3 Fig. Analyses of adult and neonatal mice.

(A) For adult, weights of WT versus mice inheriting the maternal ΔCCDIvr allele (top panels) and WT versus mice inheriting the paternal ΔCCDIvr allele (bottom panels) (Two-tailed Welch’s t-test: *p < 0.05, **p < 0.01, ***p < 0.001; error bars represent SD). (B) Allelic Igf2 expression in adult TA from mice inheriting the ΔIvr allele. Igf2 RT-PCR products were digested with MluCI to discriminate the parental B (WT B, ΔIvr, ΔCCDIvr) and C alleles. (C) and (D) All data is presented on WashU Epigenome Browser mouse mm10: chr7:142,300,000–142,750,000. (C) From top to bottom, the following are depicted: Genes, regulatory elements (SME, ICR, CCD) and probes are annotated (Sheet C in S2 Table). WashU Epigenome Browser CTCF ChIP Seq tracks (Sheet D in S2 Table). Below are normalized Capture-C interaction frequencies of F1 hybrid neonatal SkM from the viewpoint of anchors at CCD and at Igf2prox (Sheet F in S5 Table). CCD and Igf2prox anchored regions are highlighted (purple bars). Black tracks designate interactions from anchor (CCD) without an available polymorphism to discriminate alleles. Pink and blue tracks designate maternal and paternal alleles, respectively, and darker intensities correspond to stronger interactions. The biallelic interaction between sequence adjacent to CCD and Igf2prox occurs on the wild-type allele but is absent from ΔCCDIvr alleles in neonatal SkM. Normalized Capture C tracks to/from Igf2prox and CCD anchors (purple bars) on wild-type (WT, WT_maternal or WT_paternal) and ΔCCDIvr (MKO_maternal and PKO_paternal) alleles in neonatal SkM. Probe and regulatory regions are also annotated below the Capture C data.

(PDF)

S4 Fig. Igf2 promoter interactions, PaSME interactions and enhancer marks in neonatal (and adult) skeletal muscle.

(A) and (B) All data is presented on WashU Epigenome Browser mouse mm10: chr7:142,300,000–142,750,000. Presented from top: annotated genes, regulatory elements and probes and WashU Epigenome Browser CTCF ChIP Seq track (Sheet A and C in S2 Table and Sheet A in S4 Table). Normalized allele-specific Capture-C peaks on WT, MKO and PKO alleles. (A) Captured interactions from anchor at Igf2 promoter. As measured on WT alleles, boxed regions depict paternal specific interaction of Igf2 promoter to Nctc1/SME region (green bar) (Sheet G in Table 5). (B) Analysis of putative adult skeletal muscle specific enhancer (PaSME) region (green bar) in embryonic and neonatal limb (Sheet A in S2 Table, mm10 chr7:142599494–142601493). All data presented on WashU Epigenome Browser mouse mm10 and as referenced in Sheet A in S2 Table. ENCODE DNase-seq (DNase Hypersensitivity Site, DHS) data from neonatal heart and liver and embryonic limb, and H3K27ac and H3K4me2 ChIP-seq peaks in embryonic limb are shown. Below, captured interactions from anchor at PaSME in neonatal and adult skeletal muscle. On WT alleles, boxed regions depict paternal specific interaction to Nctc1/SME region (‘a’ significant, ‘b’ trending, Sheets H and I in S5 Table). On adult PKO allele, a boxed region at Igf2 promoter indicates significant enhancer-promoter (E-P) interactions (Sheet J in S5 Table). SME and PaSME regions are highlighted with green bars and Igf2 promoter region is highlighted with a blue bar. (A) and (B) Wald statistics using two methods to identify peaks are presented (* or # p < 0.05, ** or ## p < 0.01, Sheets G, H and J in S5 Table).

(PDF)

S5 Fig. Conservation of CTCF binding and chromatin structure at human IGF2/H19 locus.

(A) 3D and 2D chromatin structure at human IGF2/H19 locus (hg19 chr11:1,645,000–2,225,000). Source of publicly available data is presented in Sheet B in S2 Table. Juicebox software 3D architecture data [chromatin interaction analysis by paired-end tag sequencing (ChIA-PET)] in human GM12878 cells. The conserved interaction between CCD and proximal Igf2 CTCF sites is presented in dashed boxes. RefSeq Genes, CCD, CTCF ChIP-seq peaks (GM12878 cells, ENCODE) are noted to left and on top of Juicebox interaction map. Below are UCSC genome and WashU browser views of genes within hg19 chr11:1645000–2225000. Location of ICR, CCD, conserved regions hCR1 and hCR2 (upper panel) are designated (Sheet C in S2 Table). CTCF site polarity (green forward arrow/bars for forward, and red reverse arrow/bars for reverse) and CTCF and RAD21 ChIP seq data for indicated cell lines are presented in upper panel (Sheets B and E in S2 Table). In lower UCSC genome browser view, with annotated mouse homologous SME and PaSME regions (green bars), DNase-Seq (DHS) and ChIP-Seq (H3K27ac, H3K4me2, H3K4me1) tracks from designated adult tissues are presented. The Juicebox and UCSC and WashU browser genomes are aligned with each other. (B) CTCF is not bound to originally annotated CTCF binding site (BS) CTS5 in human cell lines; CTCF binding is shown at a newly annotated CTCF BS ICR7 (Sheet E in S2 Table).

(PDF)

S1 Table. Sheet A_Ivr 5’RACE-PCR clones; Sheet B_Ivr Ex2_Ex3 Splice Junction.

(XLSX)

S2 Table. Sheet A_Mouse Public Data; Sheet B_Human Public Data; Sheet C_Genome Coordinates for ROI; Sheet D_PWM Mouse CTCF ChIPSeq Motifs; Sheet E_PWM Human CTCF ChIPSeq_Motifs.

(XLSX)

S3 Table. Sheet A_Southern blot probes; Sheet B_Genotyping primers and pcr conditions; Sheet C_Allele-specific expression primers and RFLP assays; Sheet D_Total RNA (RT-qPCR) primers.

(XLSX)

S4 Table. Sheet A_Capture-C Probes (oligonucleotides); Sheet B_Nuclei processed for CapC; Sheet C_CapC SNPs; Sheet D_Preprocessing Stat; Sheet E_Reporter Reads.

(XLSX)

S5 Table. Sheets A-J_Statistics for designated figure sample comparison and CapC anchor; Sheet K_Neonatal Comparison Sample Summary; Sheet L_Adult Comparison Sample Summary.

(XLSX)

S1 Data. Raw Images.

(PDF)

S2 Data. Graph data and statistics.

(XLSX)

Data Availability

The accession number for raw and processed Capture-C sequencing data reported in this paper is publicly available and is deposited at GEO:GSE202669.

Funding Statement

This work was supported by grants from the National Institutes of General Medicine (https://www.nigms.nih.gov/) under award numbers GM-051279-28 to M.S.B and 5T32HD083185-02 to A.M.J, and the National Institute of Child Health and Human Development (https://www.nichd.nih.gov/) under award number F31 HD095583-01A1 to A.M.J. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.