Genome-wide binding and mechanistic analyses of Smchd1-mediated epigenetic regulation

Significance

Structural maintenance of chromosomes flexible hinge domain containing 1 (Smchd1) is a protein that plays an important role in maintaining gene silencing in many biological circumstances, including facioscapulohumeral muscular dystrophy; however, how it brings about gene silencing is unknown. Understanding the molecular mechanism by which Smchd1 contributes to stable transcriptional silencing is critical to appreciate how it functions in normal biology and when it is mutated in facioscapulohumeral muscular dystrophy. This study reveals, for the first time to our knowledge, where Smchd1 binds genome-wide, its hitherto unappreciated functional interaction with chromatin organizer CCCTC-binding factor in gene regulation, and which part of the protein is required for chromatin binding. These data lead to a new model of Smchd1 function, where it directly binds DNA to mediate 3D chromatin architecture.

Abstract

Structural maintenance of chromosomes flexible hinge domain containing 1 (Smchd1) is an epigenetic repressor with described roles in X inactivation and genomic imprinting, but Smchd1 is also critically involved in the pathogenesis of facioscapulohumeral dystrophy. The underlying molecular mechanism by which Smchd1 functions in these instances remains unknown. Our genome-wide transcriptional and epigenetic analyses show that Smchd1 binds cis-regulatory elements, many of which coincide with CCCTC-binding factor (Ctcf) binding sites, for example, the clustered protocadherin (Pcdh) genes, where we show Smchd1 and Ctcf act in opposing ways. We provide biochemical and biophysical evidence that Smchd1–chromatin interactions are established through the homodimeric hinge domain of Smchd1 and, intriguingly, that the hinge domain also has the capacity to bind DNA and RNA. Our results suggest Smchd1 imparts epigenetic regulation via physical association with chromatin, which may antagonize Ctcf-facilitated chromatin interactions, resulting in coordinated transcriptional control.

Structural maintenance of chromosomes flexible hinge domain containing 1 (Smchd1) is an epigenetic repressor that has been shown to play an essential role in autosomal and X-linked gene repression, with critical consequences for normal biology and disease. Smchd1 was originally identified as an epigenetic modifier in an N-ethyl-N-nitrosourea (ENU) mutagenesis screen (1, 2). The ENU-induced nonsense mutation, termed MommeD1, leads to dramatic reduction of Smchd1 transcripts and effectively produces a null allele (2). Mice homozygous for this allele display female-specific embryonic lethality due to failure of X chromosome inactivation (2). Although CpG island (CGI) hypomethylation was observed at promoters of a subset of X-linked genes in the absence of Smchd1 (2, 3), Smchd1-dependent gene silencing does not seem to be solely mediated by DNA methylation (4). Indeed, a study of X inactivation in human cells has suggested that SMCHD1 may provide a link between different repressive histone modifications that facilitate heterochromatin formation (5).

In addition to its role in X inactivation, Smchd1 regulates genomic imprinting of a subset of genes within the small nuclear ribonuclear protein N (Snrpn)- and Igf2r-imprinted clusters (4, 6, 7), and Smchd1/SMCHD1 is involved in regulating the expression of the clustered protocadherin (Pcdh) genes (4, 6–8). SMCHD1 has recently been implicated in the pathogenesis of facioscapulohumeral muscular dystrophy, where SMCHD1 is critical for epigenetic repression of the disease causal gene DUX4 (9, 10). Furthermore, Smchd1 deficiency has been associated with accelerated tumorigenesis in mouse models (6). These findings highlight that Smchd1/SMCHD1 participates in epigenetic regulation at multiple loci in many different cellular scenarios. However, the precise means by which Smchd1/SMCHD1 modulates gene expression in any of these cases is unknown.

Smchd1 is a noncanonical member of the structural maintenance of chromosomes (SMC) family, comprising an N-terminal ATPase domain and a C-terminal SMC hinge domain (2, 7, 11, 12). SMC proteins form dedicated complexes that play fundamental roles in chromosome dynamics and are implicated in gene regulation, DNA repair, and disease (13–15). The SMC hinge domain mediates dimerization of SMC proteins and confers differential DNA binding properties (14, 16–21). Unlike other SMC proteins, which form a composite ATPase domain from two subdomains that are brought into proximity upon SMC protein heterodimerization, Smchd1 possesses a predicted N-terminal ATPase domain encoded within a single Smchd1 protein. Interestingly, proteins containing an ATPase domain homologous to the ATPase domain of Smchd1 have recently been implicated in heterochromatin compaction and gene silencing in Arabidopsis (12, 22).

Previous results from immunofluorescence and ChIP experiments indicated that Smchd1/SMCHD1 is recruited to chromatin, including regions on the inactive X chromosome and at the DUX4 locus (2, 3, 5, 9). However, the exact site to which Smchd1/SMCHD1 is bound was not delineated, and detailed high-resolution analysis of Smchd1/SMCHD1 occupancy on a genome-wide scale was not available.

To address how Smchd1 affects gene expression at the molecular level, we use genome-wide approaches here and assess Smchd1’s chromatin occupancy in conjunction with global gene expression analysis and epigenetic profiling. We find that many Smchd1 binding sites overlap with CCCTC-binding factor (Ctcf) occupancy at both promoters and distal cis-regulatory elements. In particular, we demonstrate that Smchd1 and Ctcf display opposing functional effects in regulating the clustered Pcdh genes. To investigate how Smchd1 binds to chromatin, we have used a suite of biochemical and biophysical assays, and report that the hinge domain of Smchd1 is capable of direct DNA and RNA binding in vitro. These results indicate that like Ctcf, Smchd1 directly interacts with chromatin but exerts distinct effects on transcription, adding a further layer of complexity in chromatin dynamics and epigenetic regulation.

Results

Transcriptome Analysis of Smchd1-Null Neural Stem Cells.

To identify differentially expressed (DE) genes in neural stem cells (NSCs), we derived WT and Smchd1-null (Smchd1^{MommeD1/MommeD1}) NSCs from embryonic day 14.5 (E14.5) mouse brains. Only male cell lines were used in this study because Smchd1-null female embryos die around E10.5. We extracted RNA from NSCs (n = 3 per genotype) and performed next-generation sequencing, followed by bioinformatic analyses, to quantify the relative transcript levels. We identified 998 up-regulated and 199 down-regulated genes in Smchd1-null cells (Fig. 1A and Dataset S1). We performed a gene ontology (GO) analysis of these DE genes using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) (23, 24), and found significant up-regulation of genes involved in cell adhesion and other neural-associated functions (Table S1). As expected, the transcript level of Smchd1 was dramatically reduced in Smchd1-null NSCs, owing to nonsense-mediated decay caused by the MommeD1 mutation. Consistent with previous studies, imprinted genes, including necdin (Ndn), makorin 3 (Mkrn3), and paternally expressed gene 12 (Peg12), were significantly up-regulated in mutant cells [Fig. 1A and Dataset S1; P < 0.0001 by a rotation gene set test (ROAST) (25) of all imprinted genes]. We also found down-regulation of Grb10, for the first time to our knowledge, a gene that shows reciprocal imprinting in different areas of the brain (Dataset S1). Strikingly, almost all of the Pcdh genes within the alpha and beta clusters, a total of 31 genes, displayed elevated transcript levels (Fig. 1A and Dataset S1; P < 0.005 by a ROAST (25) of all Pcdh genes), demonstrating a more widespread effect of Smchd1 on these clusters than previously reported in studies performed in whole brain or embryos (4, 7). These differences also likely explain the striking up-regulation of GO terms associated with cell adhesion.

Fig. 1.

Gene expression analyses of WT and Smchd1-null NSCs. (A) Transcriptome-sequencing data represented by an log ratios versus mean averages (MA) plot showing the log-fold change (LogFC) of normalized expression levels between Smchd1-null and WT NSCs (n = 3 per genotype) against the average expression in log counts per million (LogCPM). Black dots are non-DE genes. Up-regulated and down-regulated genes (FC > 1.5, adjusted P < 0.01) are shown in pink and orange, respectively. Among the up-regulated genes, genes imprinted with Ndn, Mkrn3, and Peg12 are shown in blue. Pcdh alpha and beta genes are shown in red. (B) Quantitative RT-PCR (qRT-PCR) quantification of mRNA levels of Pcdh genes in alpha, beta, and gamma clusters from WT (blue) and Smchd1 null (red) NSCs (n = 3 per genotype). The qRT-PCR signal was normalized relative to the qRT-PCR signal of Rala and plotted relative to the corresponding WT sample. Data are displayed as mean + SEM and were analyzed by an unpaired two-tailed Student’s t test with a Benjamini–Hochberg correction for multiple testing. *P < 0.05; **P < 0.01; ***P < 0.001.

Table S1.

DAVID GO term analysis

Category	Term	No. DE from this category	Percentage of genes that are DE of those genes categorized by DAVID	P value	Benjamini-corrected P value
GOTERM_BP_FAT	Homophilic cell adhesion	40	4.1	1.9E-23	4.3E-20
GOTERM_BP_FAT	Biological adhesion	83	8.6	6E-21	4.5E-18
GOTERM_BP_FAT	Cell adhesion	83	8.6	5.3E-21	6E-18
GOTERM_BP_FAT	Cell–cell adhesion	49	5.1	1.8E-18	1E-15
GOTERM_BP_FAT	Ion transport	76	7.9	1.8E-11	8.2E-09
GOTERM_BP_FAT	Metal ion transport	56	5.8	2.2E-11	8.3E-09
GOTERM_BP_FAT	Transmission of nerve impulse	37	3.8	8.1E-11	0.000000026
GOTERM_BP_FAT	Monovalent inorganic cation transport	42	4.3	7E-10	0.0000002
GOTERM_BP_FAT	Potassium ion transport	29	3	1.2E-09	0.0000003
GOTERM_BP_FAT	Cation transport	57	5.9	2.5E-09	0.00000057
GOTERM_BP_FAT	Synaptic transmission	29	3	0.000000014	0.0000029
GOTERM_BP_FAT	Neuron projection development	32	3.3	0.000000027	0.0000051
GOTERM_BP_FAT	Neuron projection morphogenesis	28	2.9	0.000000043	0.0000076
GOTERM_BP_FAT	Cell projection morphogenesis	30	3.1	0.000000062	0.00001
GOTERM_BP_FAT	Cell morphogenesis involved in neuron differentiation	28	2.9	0.000000089	0.000013
GOTERM_BP_FAT	Cell part morphogenesis	30	3.1	0.00000018	0.000025
GOTERM_BP_FAT	Neuron development	36	3.7	0.00000027	0.000036
GOTERM_BP_FAT	Axonogenesis	25	2.6	0.00000053	0.000067
GOTERM_BP_FAT	Sodium ion transport	21	2.2	0.0000012	0.00015
GOTERM_BP_FAT	Neuron differentiation	42	4.3	0.0000016	0.00019
GOTERM_BP_FAT	Cell morphogenesis involved in differentiation	28	2.9	0.000002	0.00021
GOTERM_BP_FAT	Cell projection organization	36	3.7	0.0000022	0.00023
GOTERM_BP_FAT	Cell morphogenesis	33	3.4	0.000019	0.0018
GOTERM_BP_FAT	Extracellular structure organization	21	2.2	0.000019	0.0019
GOTERM_BP_FAT	Cell–cell signaling	31	3.2	0.000037	0.0032

Database for Annotation, Visualization, and Integrated Discovery (DAVID) analysis of genes significantly up-regulated in Smchd1-null NSCs compared with WT controls. Categories with fewer than 20 genes and a Benjamini P value >0.01 are not shown. No categories were significantly enriched (Benjamini P value <0.01) in the genes down-regulated in Smchd1-null NSCs.

Expression of Pcdh Genes Is Altered in Smchd1-Null NSCs.

The observed up-regulation of Pcdh genes in Smchd1-null NSCs was verified by quantitative RT-PCR experiments. In agreement with the global analysis, the majority of Pcdh genes within the alpha and beta clusters were substantially up-regulated, although expression of genes within the gamma cluster was less perturbed in Smchd1 mutants (Fig. 1B). These data suggest that Smchd1 regulates the expression of Pcdh genes, but in a nonuniform manner across the three Pcdh gene clusters.

Genome-Wide Occupancy of Smchd1 and Epigenetic Modification Profiling.

To uncover the mechanism by which Smchd1 modulates gene expression, we performed ChIP-sequencing (ChIP-seq) with an anti-Smchd1 antibody, using Smchd1-null NSCs as a negative control. Model-based analysis for ChIP-Seq 2 (MACS2) peak calling (26) identified 227 highly specific Smchd1 binding sites across the genome (Dataset S1). In parallel, we profiled key genome-wide epigenetic marks in WT and Smchd1-null NSCs (H3K4me3 and H3K27me3 ChIP-seq) and assessed CpG methylation via methyl binding domain (MBD)-seq.

Chromatin Association of Smchd1 and Epigenetic Marks at the Pcdh Gene Clusters.

The strong altered expression profiles of the clustered Pcdh genes prompted us to examine whether Smchd1 was bound at the Pcdh clusters. Initially, we identified a prominent Smchd1 peak (chr18: 37218151–37218474, fold enrichment = 5.07; P < 5 × 10⁻⁴) at the HS5-1 site located between the alpha and beta clusters (Fig. 2A and Dataset S1). HS5-1 refers to the cis-regulatory DNaseI hypersensitive site, which acts as a transcriptional enhancer of Pcdh alpha genes in the nervous system and is required for their repression in nonneuronal lineages (27, 28). We noticed that subtle Smchd1 peaks might not be detected by our stringent genome-wide analysis because our Smchd1 ChIP-seq had considerable background. When we focused on the region spanning the clustered Pcdh genes and relaxed the cutoff set for MACS2 peak calling to P < 5 × 10⁻³, four additional Smchd1 peaks were recovered, including three at promoter regions of the Pcdh genes and one at the HS5-1 site (Fig. 2A and Dataset S1). Previous studies have demonstrated that the Ctcf/Cohesin complex (which contains a Smc1/Smc3 heterodimer) is localized at the HS5-1 site, as well as promoters of actively transcribed Pcdh alpha genes (27, 29, 30). Smchd1 binding at the HS5-1 site and the Pcdh alpha promoter region was reminiscent of Ctcf/Cohesin occupancy compared with Ctcf peaks from the Encyclopedia of DNA Elements (ENCODE) E14.5 mouse brain dataset. In particular, two Ctcf/Cohesin binding sites within the HS5-1 element, HS5-1a and HS5-1b (29, 30), were both occupied by Smchd1 (Fig. 2A and Dataset S1). However, we did not detect significant Smchd1 enrichment within the region spanning the beta and gamma clusters (Fig. S1).

Fig. 2.

Aligned ChIP-seq and MBD-seq tracks showing localization of Smchd1 (blue), H3K4me3 (orange), H3K27me3 (green), and CpG methylation (pink) in WT and Smchd1-null NSCs at the Pcdh alpha cluster (A) and the Snrpn cluster (B). Sequenced reads from two (ChIP-seq) or three (MBD-seq) biological replicates of each genotype were combined and plotted as normalized read coverage on the y axis against the genomic location along the horizontal axis. Smchd1 peaks identified by MACS2 are annotated as follows: **P < 5 × 10⁻⁴, q < 0.1; *P < 5 × 10⁻³. (Inset) Zoomed-in view of Smchd1 peaks. Smchd1 peak locations are shown using blue lines that extend through all tracks. Ctcf peaks from the ENCODE E14.5 mouse brain dataset [University of California, Santa Cruz (UCSC) accession no. wgEncodeEM002595] are represented as gray bars below the tracks. The positions of Ctcf peaks at HS5-1a (chr18: 37217137–37217798) and HS5-1b (chr18: 37218262–37218753), as previously published by Monahan et al. (29), were converted from mm9 to mm10 mouse genome and are marked by the black triangles.

Fig. S1.

Aligned ChIP-seq or MBD-seq tracks showing localization of Smchd1 (blue), H3K4me3 (orange), H3K27me3 (green), and CpG methylation (pink) in WT and Smchd1-null NSCs at the Pcdh alpha, beta, and gamma clusters. Sequenced reads from replicates of each genotype were combined and plotted as normalized read coverage on the y axis against the genomic location along the horizontal axis. Smchd1 peaks are marked with blue bars and with asterisks (*P < 5 × 10⁻³; **P < 5 × 10⁻⁴). Ctcf peaks from the ENCODE E14.5 mouse brain dataset (UCSC accession no. wgEncodeEM002595) are represented as gray bars below the tracks.

We assessed whether Smchd1 deficiency altered the epigenetic modifications at the Pcdh clusters. We noticed an acquisition of the active mark H3K4me3 and a reduction of repressive CpG methylation at the promoters of individual Pcdh genes in Smchd1-null cells, predominantly for Pcdha1-a12 (Fig. 2A and Fig. S1). This result was confirmed by read quantification comparing the Smchd1-null and WT NSC data. The calculated log₂-fold change for H3K4me3 and CpG methylation at the region encompassing Pcdha1-a12 promoters was ∼2.23 and −0.74, respectively (Table S2). Similar analysis also revealed a discernible increase in H3K27me3 abundance in Smchd1-null cells at this region (Table S2). Because H3K27me3 is usually associated with gene silencing, this finding was unexpected, potentially indicating some form of epigenetic compensation, albeit inadequate to maintain Pcdh repression. At the beta and gamma clusters, differential enrichment for H3K4me3 and CGI methylation was less apparent and H3K27me3 was not significantly changed in Smchd1-null NSCs (Fig. S1 and Table S2).

Table S2.

Quantification of H3K4me3, H3K27me3, and CGI methylation

Region	Chromosome	Start	End	Log-fold change	P value
H3K4me3 ChIP-seq
Pcdh alpha cluster	18	36925285	37025230	2.23945752	0.000882384
Pcdh beta cluster	18	37259998	37518652	1.013591647	0.08767658
Pcdh gamma cluster	18	37656945	37824546	0.178673365	0.750018041
Ndn	7	62343277	62353277	0.969333241	0.169188606
Mkrn3	7	62415139	62425139	0.538511874	0.294488622
Peg12	7	62459510	62469510	0.97873133	0.066078654
H3K27me3 ChIP-seq
Pcdh alpha cluster	18	36925285	37025230	0.98067961	1.44E-07
Pcdh beta cluster	18	37259998	37518652	−0.035730436	0.57755971
Pcdh gamma cluster	18	37656945	37824546	−0.108575033	0.031034283
Ndn	7	62343277	62353277	0.122207209	0.211780517
Mkrn3	7	62415139	62425139	−0.088070464	0.412371146
Peg12	7	62459510	62469510	−0.253657027	0.150760986
MBD-seq
Pcdh alpha cluster	18	36925285	37025230	−0.739769365	0.039644095
Pcdh beta cluster	18	37259998	37518652	0.321751021	0.32336855
Pcdh gamma cluster	18	37656945	37824546	0.308676272	0.156179791
Ndn	7	62343277	62353277	−0.441461563	0.611197955
Mkrn3	7	62415139	62425139	−2.402747677	0.004974309
Peg12	7	62459510	62469510	−2.429274144	0.611197955

Relative number of reads mapped within each region (normalized against the total number of reads across the genome) in Smchd1-null compared with WT NSCs in ChIP-seq or MBD-seq experiments.

These data demonstrate that Smchd1 is physically associated with regulatory elements that are implicated in Ctcf/Cohesin-mediated transcription of Pcdh alpha genes. In Smchd1-null NSCs, altered epigenetic modifications, featuring increased H3K4me3 and reduced CpG methylation at promoters, are concomitant with the induced expression, mostly evident for the Pcdh alpha genes.

Chromatin Association of Smchd1 and Epigenetic Marks at the Snrpn-Imprinted Gene Cluster.

Smchd1-null cells show failed genomic imprinting at the Snrpn cluster, specifically for the distal half of the cluster encoding Ndn; melanoma antigen, family L, 2 (Magel2); Mkrn3; and Peg12, which are biallelically expressed in Smchd1-null cells (4, 7). We again find up-regulation of Ndn, Mkrn3, and Peg12 here (Fig. 1A), indicative of loss-of-imprinting and biallelic expression. Analysis of chromatin marks and CpG methylation in this region showed the expected enrichment of H3K4me3 at the promoters of the up-regulated genes (Ndn, Mkrn3, and Peg12) and loss of CpG methylation. No significant change in H3K27me3 enrichment was observed (Fig. 2B and Table S2). We next examined Smchd1 localization in this region using MACS2 with higher (P < 5 × 10⁻⁴) and lower (P < 5 × 10⁻³) stringency settings, as we did for the Pcdh cluster. This analysis revealed seven Smchd1 peaks of enrichment: three within the cluster of genes and four more distally located. Interestingly, two Smchd1 peaks near Magel2 and Mkrn3, and two at the distal end of the region, overlap with Ctcf binding sites, suggesting that Smchd1 and Ctcf also have some relationship at the Snrpn locus. However, at three other Smchd1 peaks (at the Mkrn3 promoter and at two sites downstream of the genes), no such overlap exists, demonstrating that Smchd1 and Ctcf can bind unique sites (Fig. 2B).

Chromatin Association of Smchd1 and Epigenetic Marks at Homeobox (Hox) Gene Clusters.

Unexpectedly, our genome-wide analysis also identified Smchd1 binding sites at all four Hox gene clusters, some of which overlap with Ctcf peaks from the ENCODE E14.5 mouse brain dataset (Fig. S2). Hox genes encode homeodomain transcription factors that are expressed in a spatially and temporally restricted manner, regulated partly via epigenetic mechanisms (31–33). In embryonic forebrain, the region from which the NSCs were derived, Hox genes are repressed via polycomb group protein-mediated silencing (34). We did not detect expression of Hox genes in our forebrain-derived NSCs irrespective of the presence or absence of Smchd1. Furthermore, in both WT and Smchd1-null NSCs, all four Hox gene clusters were decorated with extensive H3K27me3 methylation (Fig. S2). Thus, Smchd1 binding is dispensable for Hox gene silencing in NSCs.

Fig. S2.

(A–D) Aligned ChIP-seq or MBD-seq tracks showing localization of Smchd1 (blue), H3K4me3 (orange), H3K27me3 (green), and CpG methylation (pink) in WT and Smchd1-null NSCs at the Hox cluster (HoxA–HoxD, respectively). Sequenced reads from replicates of each genotype were combined and plotted as normalized read coverage on the y axis against the genomic location along the horizontal axis. The most significant Smchd1 peaks are marked with blue bars and with asterisks (**P < 5 × 10⁻⁴, q < 0.1). Ctcf peaks from the ENCODE E14.5 mouse brain dataset (UCSC accession no. wgEncodeEM002595) are represented as gray bars below the tracks.

Smchd1 Occupancy Coincides with Ctcf Binding Sites.

To annotate the identified Smchd1 binding sites functionally across the genome, we defined the distribution of Smchd1 peaks with respect to gene transcription start sites (TSSs) using the Genomic Regions of Enrichment Analysis Tool (GREAT) algorithm (35). We noticed that only 24% of Smchd1 peaks occur within 5 kb of annotated TSSs (Fig. 3A). In contrast, there was a tendency for Smchd1 to be localized at regions distant from the TSSs (Fig. 3A). These data suggest Smchd1 is not only involved in transcriptional repression at the TSSs but may also bind more distal regulatory elements, such as enhancers or insulators. To test this possibility, we examined Smchd1 occupancy in relation to cis-regulatory elements in mouse E14.5 brain, defined by their transcription factor and chromatin mark profiles (36). Strikingly, we observed that overlap between Ctcf and Smchd1 at the Pcdh and Snrpn clusters extended genome-wide: 138 of the 227 Smchd1 binding sites overlapped with Ctcf, of which 75 were devoid of promoter or enhancer elements (Fig. 3B and Dataset S1). To rule out the possibility that those 75 sites contained inactive promoters (without H3K4me3 mark or Pol II binding), we applied the GREAT algorithm on those 75 sites and found that the majority of them were distant from TSSs (Fig. S3A). Thus, our comparative analysis indicated that genome-wide Smchd1 occupancy coincided with Ctcf binding sites, featuring both promoters that are within close proximity of the TSSs and distal enhancers and insulators. This finding was further supported by our de novo motif analysis for Smchd1, which identified two statistically significant motifs corresponding to the Ctcf consensus sequence within 119 of the Smchd1 peaks (Fig. 3C and Dataset S1). Additionally, a less common motif containing a sequence recognized by RE1 silencing transcription factor/neuron-restrictive silencer factor (Rest/Nrsf), a neuronal gene-specific transcription repressor (37, 38), was identified in 38 Smchd1 peaks (Fig. 3C and Dataset S1), whereas the other motifs identified do not represent known transcription factor binding sites. The top 10 Smchd1 motifs identified in this analysis are given in Fig. S3B.

Fig. 3.

Genome-wide analysis of Smchd1 chromatin occupancy. (A) Distribution of 227 Smchd1 binding sites relative to TSSs calculated by the GREAT algorithm. The percentages of association are plotted on the y axis, and categories of the distances between Smchd1 peaks and assigned TSSs are plotted on the x axis. (B) Venn diagram showing the overlap between 227 Smchd1 binding sites and the E14.5 mouse brain-specific cis-regulatory elements, including promoters (denoted by H3K4me3 and Pol II sites), enhancers, and insulators (Ctcf binding sites), as previously published by Shen et al. (36). The number of peaks corresponding to motif 4 (Rest/Nrsf consensus) from the de novo motif analysis is shown. Below the Venn diagram, we show the frequency with which other motifs are observed in those peaks without annotated regulatory motifs. (C) De novo motif analysis based on Smchd1 peak sequences identified motif 5 and motif 8, which are highly similar to the Ctcf consensus sequence, and motif 4, which is highly similar to the Rest/Nrsf consensus.

Fig. S3.

(A) GREAT analysis of 75 Smchd1 binding sites that overlap with Ctcf binding sites but do not contain promoters or enhancers. (B) Top 10 motifs identified in Smchd1 peak sequences using de novo motif analysis MEME. Motifs 4, 5, and 8 are shown in Fig. 3.

Smchd1 and Ctcf Mediate Opposite Effects on the Expression of Pcdh Genes.

Given the potential co-occupancy of Smchd1 and Ctcf at cis-regulatory elements, we next investigated whether Smchd1 and Ctcf could coordinately regulate gene expression, focusing on the clustered Pcdh genes. We performed shRNA-mediated knockdown of Ctcf in NSCs with two independent hairpins, validated at both the mRNA and protein levels (Fig. S4). We selected a set of 13 Pcdh genes to test, spread throughout the alpha, beta, and gamma clusters. Upon knockdown of Ctcf, there was noticeably reduced expression of a number of the Pcdh genes, compared with the negative control (Fig. 4A), similar to the down-regulation of clustered Pcdh genes observed in Ctcf-deficient mouse brains (39). We found that the up-regulated expression of several Pcdh genes in Smchd1-null NSCs was partially reversed upon Ctcf depletion (Fig. 4A). Together, these results suggest that Smchd1 and Ctcf evoke opposing effects on transcription of Pcdh genes, particularly those Pcdh genes of alpha and beta clusters. Whether this effect is coordinated through protein–protein interactions between Smchd1 and Ctcf is less clear, because we did not detect apparent interactions by immunoprecipitation under native conditions (Fig. S4).

Fig. S4.

Validation of shRNA-mediated knockdown of Ctcf and IP of Smchd1 and Ctcf. (A) Quantitative RT-PCR quantification of mRNA levels of Ctcf in WT (blue) and Smchd1-null (red) NSCs upon Ctcf knockdown by two shRNA hairpins, Ctcf.6 (dotted bar) and Ctcf.7 (dashed bar), along with a nonsilencing negative control (Nons, plain bar) (n = 3 per genotype and treatment). The qRT-PCR signal was normalized relative to the averaged signal of Gusb and Hprt, and was plotted relative to the Nons WT sample. Data are displayed as mean ± SEM. Comparisons between expression levels in Ctcf knockdown and Nons control samples were analyzed by an unpaired two-tailed Student’s t test. **P < 0.01. (B) Western blot analysis of protein levels in Nons, Ctcf.6, and Ctcf.7 shRNA-transduced WT and Smchd1-null NSCs. Whole-cell extract was prepared and quantified. An equivalent amount of total protein was analyzed with antibodies against Smchd1 (Top), Ctcf (Middle), and tubulin (Bottom) as the loading control. The asterisk denotes the nonspecific signals produced by the antibody. (C) Smchd1 IP and Ctcf IP were performed with the anti-Smchd1 and anti-Ctcf antibodies, respectively, in WT NSC whole-cell extracts. Control IP was performed with the rabbit IgG. Western blot analysis of whole-cell extracts and immunoprecipitated samples was performed with antibodies against Smchd1 (Top) and Ctcf (Bottom), showing the successful IP of Smchd1 or Ctcf, but not co-IP of the two proteins. The asterisk denotes the nonspecific signals produced by the antibody.

Fig. 4.

Smchd1 and Ctcf play opposing roles in regulating Pcdh gene expression. (A) qRT-PCR quantification of mRNA levels of a subset of Pcdh genes in WT (blue) and Smchd1 null (red) NSCs with Ctcf knockdown by two validated shRNAs, Ctcf.6 (dotted bar) and Ctcf.7 (dashed bar), and the nonsilencing negative (Nons; plain bar) control (n = 3 per genotype and treatment). The qRT-PCR signal was normalized relative to the averaged signal of Gusb and Hprt, and plotted relative to the corresponding WT samples. Data are displayed as mean + SEM. Comparison between expression levels in Ctcf knockdown and Nons control samples from Smchd1-null NSCs was analyzed by an unpaired two-tailed Student’s t test. *P < 0.05; **P < 0.01; ***P < 0.001. (B) Aligned ChIP-seq tracks showing localization of Ctcf at the Pcdh alpha cluster. ChIP-seq experiments were performed in untransduced (Untrans) Smchd1-null and WT NSCs, Nons control, or Ctcf knockdown (Ctcf KD) transduced WT NSCs. Sequenced reads from two replicates of each genotype and treatment were combined and plotted as normalized read coverage on the y axis against the genomic location along the horizontal axis. Ctcf peaks from the ENCODE E14.5 mouse brain dataset (UCSC accession no. wgEncodeEM002595) are represented as gray bars below the tracks.

To elucidate the observed opposing effects of Smchd1 and Ctcf, we compared the chromatin localization of Ctcf in WT and Smchd1-null cells by performing Ctcf ChIP-seq, using NSCs with Ctcf knockdown as the negative control. Intriguingly, we identified additional Ctcf peaks at the promoters of Pcdha1-a12 genes in Smchd1-null cells (Fig. 4B and Dataset S1), concurrent with the gain of H3K4me3 and loss of CpG methylation as shown in Fig. 2A. These data are highly suggestive that Smchd1 binding is essential for stabilizing a repressive chromatin environment and potentially involved in antagonizing Ctcf binding at the Pcdh alpha promoters, resulting in coordinated regulation.

Smchd1 Hinge Domain Binds to DNA in Vitro.

Although our ChIP experiments successfully captured Smchd1–chromatin interactions, it was unclear how Smchd1 bound to chromatin. Given that the hinge domains of other SMC proteins have been implicated in DNA binding (14, 16–19, 21), we produced a recombinant Smchd1 hinge domain to test whether it too bound to DNA to direct Smchd1’s interaction with chromatin (Fig. 5A). In parallel, we generated a mutant Smchd1 hinge domain with a single amino acid substitution, R1867G (Fig. 5A), mimicking the mutation present in a facioscapulohumeral muscular dystrophy type 2 family caused by a missense SMCHD1 mutation (40). EMSA with ssDNA demonstrated a shift of 15-mer poly-dT and poly-dC by the WT Smchd1 hinge domain in a concentration-dependent manner, indicating binding, but there was no binding of poly-dA ssDNA (Fig. 5B). The highly positive charge of the Smchd1 hinge domain meant it remained in the wells, or even ran toward the anode on these gels, similar to the pattern observed for the hinge domains of other SMC proteins (18). Binding activity was dramatically diminished with the pathogenic mutation (Fig. 5B).

Fig. 5.

Hinge domain of Smchd1 binds to DNA in vitro. (A) Coomassie-stained 4–12% (wt/vol) Bis-Tris SDS/PAGE gel of purified WT recombinant Smchd1 hinge domain (WT) and mutated protein with the R1867G substitution. (B) EMSA with 50 nM 15-mer single-stranded poly-dT (Left), poly-dC (Middle), and poly-dA (Right) using protein at increasing concentrations as indicated. (C) Thermostability of the WT (Top) and R1867G mutant (Bottom) Smchd1 hinge domain measured in the presence or absence of ssDNA 30-mer poly-dT. Binding with oligonucleotides increased the stability of the WT protein, thus inducing a shift in the melting temperature. The plot is representative of two independent experiments. (D and E) The c(s) distribution, plotted as a function of Svedberg (S) and detected at 488 nm with the ratio of protein/oligonucleotide concentration as indicated in individual plots. (D) Profiles of c(s) distribution for the Smchd1 hinge domain WT (Top) and R1867G mutant (Bottom) with sense ssDNA unmethylated oligonucleotide corresponding to the HS5-1b Smchd1/Ctcf binding site. (E) Profiles of c(s) distribution for the Smchd1 hinge domain WT bound to sense ssDNA unmethylated oligonucleotide corresponding to the Smchd1/Ctcf binding site near Magel2 in the Snrpn cluster (Top) or the methylated equivalent (Bottom). (F) Summary of K_d values of the WT and R1867G (RG) mutant Smchd1 hinge domain binding to the listed oligonucleotides. The ssDNA containing Ctcf consensus sequence in either sense or antisense orientation and dsDNA produced by annealing the two oligonucleotides were tested individually. Oligonucleotides with an unmethylated or methylated Cyt residue (highlighted in red) were tested separately. The K_d was calculated based on at least three independent experiments and is displayed as mean ± SE.

To verify this finding, we adopted the thermal shift assay (TSA), which monitors the thermal denaturation of sample protein by the fluorescence intensity of the dye SYPRO orange upon its binding to denatured protein (41, 42). In agreement with the EMSA results, we observed an elevated melting temperature of the WT Smchd1 hinge domain in the presence of ssDNA 30-mer poly-dT, indicating increased thermal stability induced by oligonucleotide binding. Furthermore, although the melting temperature of the R1867G mutant was indicative of a stable and correctly folded protein, such a shift was not observed for the R1867G mutant in the presence of oligonucleotides, reflecting its reduced capacity to bind DNA (Fig. 5C).

To assess the Smchd1 hinge domain–oligonucleotide interactions quantitatively, we performed analytical ultracentrifugation (AUC). Using sedimentation velocity experiments and analyzing the data via the continuous size [c(s)] distribution or species analysis method (as implemented in the program SEDFIT), the concentration of bound 6-carboxyfluorescein (6-FAM) fluorescently labeled oligonucleotide in relation to the known concentration of the free oligonucleotide and protein was determined. From this experiment, the dissociation constant (K_d) was calculated (details are provided in SI Methods and Fig. S5). This method has the advantage of also reporting the oligomeric state of the free protein and the protein/oligonucleotide complex. Within the concentration range of 1–40 μM, both the WT and R1867G mutant Smchd1 hinge domain proteins are dimeric, either free in solution or in complex with the oligonucleotides tested here.

Fig. S5.

(A) Distribution profiles for the Smchd1 hinge domain R1867G mutant with unmethylated 20-mer dsDNA (annealed with ssDNA containing the Ctcf consensus sequence in sense and antisense orientations) containing either one or two 5′ 6-FAM molecules. The c(s) distribution (normalized to one to account for the differences in absorbance between one and two 6-FAM molecules per dsDNA) was plotted as a function of Svedberg (S) detected at 488 nm with an 8:1 ratio of protein to DNA. The distribution profiles illustrate that, as expected, the ratio of bound vs. unbound DNA remains constant regardless of the presence of either one or two 6-FAM molecules, further demonstrating that increasing the number of 6-FAM molecules does not affect affinity, as evidenced by the lack of a peak for the DNA/protein complex. (B) Continuous mass distribution for the Smchd1 hinge domain WT protein (24 μM) with ssDNA HS5-1b sense unmethylated oligonucleotide (1 μM). The distribution shows a peak mass at 56.9 kDa, consistent with one oligonucleotide bound per protein dimer. The fitted frictional ratio was 1.40, although as noted in SI Methods, we found this ratio to be unstable, fitting to values between 1.29 and 1.40. The statistics for this fit to the sedimentation velocity data are as follows: rmsd = 0.0039, runs test Z score = 3.23. (C–I) Profiles of c(s) distribution for the Smchd1 hinge domain WT (Left) and R1867G mutant (Right) with oligonucleotides. The c(s) distribution is shown, plotted as a function of Svedberg detected at 488 nm, with the ratio of protein/oligonucleotide concentration as indicated in individual plots. (C) Fifteen-mer ssDNA poly-dA. (D) Fifteen-mer ssDNA poly-dC. (E) Methylated 20-mer ssDNA containing Ctcf consensus sequence from Pcdh HS5-1b in sense orientation. (F) Unmethylated 20-mer ssDNA containing Ctcf consensus sequence in antisense orientation. (G) Methylated 20-mer ssDNA containing Ctcf consensus sequence in antisense orientation. (H) Unmethylated 20-mer dsDNA annealed with ssDNA in Fig. 5D and in G. (I) Methylated 20-mer dsDNA annealed with ssDNA in E and G.

Consistent with the EMSA results, poly-dC displayed higher binding affinity than poly-dA. This result can be seen qualitatively when comparing the c(s) distributions of WT Smchd1 in the presence of either poly-dC or poly-dA (Fig. S5 C and D). The poly-dC oligonucleotide distribution (Fig. S5D) clearly shows a peak at ∼3.5 Svedberg, consistent with the Smchd1/oligonucleotide complex, whereas this peak is absent in the distribution with poly-dA (Fig. S5C). The K_d for poly-dC was determined to be 2.5 ± 0.2 μM, and the K_d for poly-dA was 66 ± 9 μM. The weak binding evident in the poly-A continuous size distribution and EMSA experiments also demonstrates that the 6-FAM label does not significantly contribute to protein binding. As expected based on EMSA and TSA data (above), the binding affinity was substantially decreased for the R1867G mutant hinge domain with both oligonucleotides (Fig. S5 C and D).

We next tested oligonucleotides corresponding to the summit of the Smchd1 peak at the HS5-1b site. We found that the K_d of 20-mer ssDNA containing a Ctcf motif in sense orientation exhibits comparable binding to 15-mer poly-dC and, again, that the R1867G mutant hinge domain had dramatically decreased binding affinity (Fig. 5 D and F). In contrast, the binding affinity for the anti–sense-orientated ssDNA was considerably compromised, with the K_d being about 10-fold higher (Fig. 5F and Fig. S5F). The dsDNA produced by annealing those two oligonucleotides was also bound by the WT hinge domain, albeit at an intermediate affinity (Fig. 5F and Fig. S5H).

Previous studies demonstrated that CpG methylation at a single site could block Ctcf binding to 20-mer dsDNA probes (30). Interestingly, similar CpG methylation did not abrogate binding of the Smchd1 hinge domain to the oligonucleotides (Fig. 5F and Fig. S5 E, G, and I). We next tested the effect of DNA methylation of two CpG sites next to a Ctcf motif within a Smchd1 binding site, between Magel2 and Mkrn3 in the Snrpn-imprinted cluster (Fig. 2B) Here, methylation resulted in enhanced affinity of the WT hinge domain for the oligonucleotide (Fig. 5 E and F). Although the exact binding mode is unclear and subject to further investigation, together, these results demonstrate that the Smchd1 hinge domain binds to DNA in vitro, and therefore is a likely candidate for recruiting Smchd1 to its chromatin binding sites.

Because the Smchd1 hinge domain displayed the highest affinity for ssDNA, we used EMSA and AUC to analyze its RNA binding capacity. We found that the WT hinge domain shifted both 15-mer polyU and polyA RNA oligonucleotides in a concentration-dependent manner in EMSA (Fig. S6A). AUC with the RNA oligonucleotides corresponding to the forward and reverse orientations of the Pcdha12 exonic Ctcf motif (29) demonstrated approximately equivalent binding as for the ssDNA oligonucleotides of the same motif (Fig. S6B). In each case, binding was compromised for the R1867G mutant protein (Fig. S6). These data raise the possibility that Smchd1, like Ctcf (43, 44), may bind both DNA and RNA moieties to achieve its epigenetic function.

Fig. S6.

Hinge domain of Smchd1 binds to RNA in vitro. (A) EMSA with 50 nM 15-mer RNA poly-U (Left) and poly-A (Right) using protein at increasing concentrations as indicated. (B) Summary of K_d values of the WT and R1867G (RG) mutant Smchd1 hinge domain binding to the ssDNA and RNA. (C–F) Continuous c(s) distribution profiles for the Smchd1 hinge domain WT (Left) and R1867G mutant (Right) with oligonucleotides. The c(s) distribution is shown, plotted as a function of Svedberg detected at 488 nm, with the ratio of protein/oligonucleotide concentrations as indicated in individual plots. (C) ssDNA Pcdha12 exon Ctcf site sense. (D) ssDNA Pcdha12 exon Ctcf site antisense. (E) RNA Pcdha12 exon Ctcf site sense. (F) RNA Pcdha12 exon Ctcf site antisense.

SI Methods

Mice.

Mice carrying the MommeD1 mutation as previously described (2) were maintained on the FVB/N inbred background, and additionally backcrossed with C57BL/6 mice for more than 15 generations to produce C57BL/6 MommeD1 congenic mice. Genotyping was as previously described (2, 7). All experimental animals were treated in accordance with the Australian Government National Health and Medical Research Council guidelines under approval from the Animal Ethics Committees of the Walter and Eliza Hall Institute (WEHI AEC 2011.027) and the Queensland Institute of Medical Research (QIMR AEC A0812-610M).

Derivation and Culture of NSCs.

Brains from E14.5 embryos derived from C57BL/6 Smchd1^MommeD1/+ males mated with FVB/N Smchd1^MommeD1/+ females were dissected out in Leibovitz’s L-15 Medium (Gibco). The meninges and subcortical tissue were removed before the isolated forebrains were transferred into HBSS (Gibco) containing DNase I (1 mg/mL; Sigma–Aldrich) and 0.5% trypsin-EDTA (Gibco) (53). Tissues were digested at 37 °C in 5% (vol/vol) CO₂ for 20 min, followed by three washes in HBSS. The tissue was then transferred and mechanically dissociated in NeuroCult NSC Basal Medium (Mouse) (StemCell Technologies) containing NeuroCult Proliferation Supplement (Mouse) (StemCell Technologies), 0.2% heparin solution (StemCell Technologies), recombinant human EGF (20 ng/mL; StemCell Technologies), and recombinant human basic FGF (20 ng/mL; StemCell Technologies). Cells were plated onto plates coated with polyornithine (15 ng/mL; Sigma) and laminin (10 ng/mL; Sigma) at a density of 200,000 cells per square centimeter. Cells were grown at 37 °C in 5% (vol/vol) CO₂ incubator and passaged every 2 d using Accutase (Sigma–Aldrich) to detach the cells from the plates.

Quantitative RT-PCR.

Quantitative RT-PCR was performed as described previously (7) using Platinum SYRB Green qPCR SuperMix-UDG (Life Technologies), PrimeTime qPCR Assays (Integrated DNA Technologies), or EvaGreen qPCR Master Mix (Biotium). Relative gene expression was calculated using the 2^−ddCt method (55) or ViiA 7 software (Life Technologies) using housekeeping genes as a control for variation in cDNA concentration between samples. Expression of genes in Smchd1^{MommeD1/MommeD1} NSCs relative to the corresponding WT sample was plotted. Sequences of all primers used are available upon request.

Retroviral-Mediated Knockdown of Ctcf in NSCs.

shRNA constructs that target Ctcf were designed using the Designer of Small Interfering RNA (DSIR) website (biodev.cea.fr/DSIR/DSIR.html) (56) and subcloned into the LTR mir30 puromycin-blue fluorescent protein (LMP-BFP) vector, a modified version of LMP-GFP (57). Retroviral supernatants were prepared by transient transfection of 293T cells in 10-cm² plates with a mix of 14.4 μg of LMP-BFP vector containing the shRNA sequence for Ctcf knockdown (Ctcf.6: TGCTGTTGACAGTGAGCGCCAGGTGCAATTGAGAACATTATAGTGAAGCCACAGATGTATAATGTTCTCAATTGCACCTGTTGCCTACTGCCTCGGA, Ctcf.7: TGCTGTTGACAGTGAGCGCCAGGTCGAAGATCAGAATACATAGTGAAGCCACAGATGTATGTATTCTGATCTTCGACCTGATGCCTACTGCCTCGGA) or nonsilencing control (TGCTGTTGACAGTGAGCGATCTCGCTTGGGCGAGAGTAAGTAGTGAAGCCACAGATGTACTTACTCTCGCCCAAGCGAGAGTGCCTACTGCCTCGGA) (58), along with 0.6 μg of pCMV-VSVg and 4.8 μg of pCMV-gag-pol vectors using the calcium phosphate precipitation method. The 293T cells were treated with 25 μM chloroquine for 30 min before transfection in DMEM with 10% (vol/vol) FCS. The precipitated DNA was added drop-wise to the cells, and the media were changed after an 8-h incubation. Media were replaced with the NSC culture media after 24 h, and viral supernatants were harvested the following day by centrifugation at 500 × g at 4 °C for 10 min. Twenty milliliters of viral supernatant was mixed with 4.25 mL of PEG 8000 at 4 °C for 1.5 h by inverting every 20 min and kept still overnight at 4 °C. The mixture was spun at 3,600 × g at 4 °C for 45 min. Supernatant was removed, and the pellet was resuspended with ∼100 μL of the NSC culture media and stored at −80 °C as 10-μL aliquots.

For retroviral transduction of NSCs, cells were seeded at a density of 45,000 cells per square meter 16 h before transduction. Viral supernatant was added to the culture together with polybrene at a final concentration of 4 μg/mL. On the next day, cells were selected with 4 μg/mL puromycin.

Transcriptome Sequencing and Analysis.

RNA was extracted from WT and Smchd1^{MommeD1/MommeD1} male NSCs using an RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions and quantified using a NanoDrop 1000 Spectrophotometer (Thermo Scientific). RNA integrity was assessed with an Agilent Bioanalyzer 2100 (Agilent Technologies) according to the manufacturer’s instructions. Sequencing libraries were prepared with a TruSeq Stranded Total RNA Sample Preparation Kit (Illumina) according to the manufacturer’s instructions. Next-generation sequencing was performed by the Australian Genome Research Facility with an Illumina HiSeq 2000 platform for 100-bp paired-end reads. Reads from each sample were mapped to the mouse genome (mm10) using the Rsubread program (version 1.10.5) (59) and summarized at the gene level using the featureCounts program (60). Subsequent analysis was carried out using edgeR (61) and limma (62) Bioconductor software. The counts were transformed into counts per million (cpm) to standardize for differences in library size, and filtering was carried out to retain genes with a baseline expression level of at least 0.5 cpm in three or more samples. Data were trimmed mean of M values (TMM)-normalized (61), and linear models using observational and sample weights (63, 64) were fitted to summarize over replicate samples. Moderated t-statistics were used to assess differential expression between Smchd1-null and WT samples, with genes ranked according to their false discovery rate (65).

MBD-Seq.

Genomic DNA was extracted from WT and Smchd1^{MommeD1/MommeD1} male NSCs using an AllPrep DNA/RNA Mini Kit (Qiagen) and quantified using a NanoDrop 1000 Spectrophotometer. DNA was fragmented using a Covaris S220 Focused-Ultrasonicator (Covaris, Inc.) with specific settings (peak power: 175, duty factor: 10, cycles per burst: 200, duration: 600 s) to yield fragments ranging from 50 to 400 bp with an average size of 200 bp, as confirmed by means of an Agilent Bioanalyzer 2100 (Agilent Technologies) according to the manufacturer’s instructions. Methylated DNA was isolated by binding to the methyl-CpG binding domain of human MBD2 protein-coupled beads using a MethylMiner Methylated DNA Enrichment Kit (Life Technologies) according to the manufacturer’s instructions. The DNA was eluted from MBD-coupled beads in two salt concentration cuts: first, a 600 mM NaCl cut to remove poorly methylated DNA, followed by a 2 M NaCl cut to elute highly methylated DNA. The highly methylated cut was then used for the preparation of libraries for next-generation sequencing. Next-generation library preparation (Illumina TruSeq DNA Sample Preparation Kit) and sequencing (HiSeq 2000 platform for 49-bp single-end reads) were performed by BGI Tech Solutions. Reads were processed and analyzed in the same manner as described below for the ChIP-seq experiments.

ChIP and Sequencing Analysis.

WT and Smchd1^{MommeD1/MommeD1} NSCs cultured in 10-cm² dishes were cross-linked with formaldehyde at a final concentration of 1% (vol/vol) for 10 min at room temperature and then quenched with Gly at a final concentration of 125 mM for 5 min at room temperature. Cells were washed twice in ice-cold PBS buffer containing 0.5 mM PMSF and scraped from the plates. Cells were pelleted by centrifugation at 400 × g at 4 °C for 5 min. Approximately 5–10 million cells were lysed with 1 mL of immunoprecipitation (IP) buffer [150 mM NaCl, 50 mM Tris⋅HCl (pH 7.5), 5 mM EDTA (pH 7.5), 0.5% (vol/vol) Nonidet P-40, 1.0% (vol/vol) Triton X-100] supplemented with protease inhibitors (cOmplete mixture tablets; Roche). Nuclei were collected by centrifugation at 12,000 × g at 4 °C for 1 min. The nuclei pellet was washed once with IP buffer and then subjected to sonication in 130 μL of IP buffer using the Covaris S220 sonicator (peak power: 125, duty factor: 10, cycle/burst: 200, duration: 720 s). An aliquot of fragmented chromatin was reverse–cross-linked, and DNA was extracted as described below. The amount of DNA was quantified for estimation of the amount of chromatin for each ChIP experiment. The chromatin was used for setting up overnight IP at 4 °C. IP samples were prepared by mixing the required amount of chromatin with IP buffer to 500 μL together with the corresponding antibody and incubated at 4 °C overnight. For Smchd1 ChIP, 60 μg of chromatin was incubated with 7.5 μg of anti-Smchd1 antibody (31865; Abcam). For H3K4me3 and H3K27me3 ChIP, 6 μg of chromatin was incubated with 1.5 μg of anti-H3K4me3 antibody (07-473; Millipore) and 3 μg of anti-H3K27me3 antibody (07-449; Millipore), respectively. For Ctcf ChIP, 15 μg of chromatin was incubated with 1.2 μL of anti-CTCF antibody (07-729; Millipore). Antibody-enriched fractions were captured by incubation with 6.25–30 μL of Dynabeads Protein G (Life Technologies) at 4 °C for 1 h. Immunoprecipitated fractions were eluted with 50 μL of elution buffer containing 20% (wt/vol) SDS and 1 M NaHCO₃ by incubation at 65 °C for 15 min, and the elution procedure was repeated once for a total of 100 μL of eluate. For the input control sample, 5 μL of fragmented chromatin was mixed with 95 μL of elution buffer for subsequent reverse–cross-linking steps. One hundred microliters of the input control sample or eluate was mixed with 100 μL of 400 mM NaCl and 4 μL of 5 mg/mL RNaseA (Sigma–Aldrich) and incubated at 65 °C for 4 h. Two microliters of 0.5 M EDTA (pH 7.5), 2 μL of 1 M Tris⋅HCl (pH 7.5), and 4 μL of 20 mg/mL Proteinase K (Qiagen) were added to the solution and incubated at 65 °C for 1 h. Samples were cooled down to room temperature, and DNA was extracted using a MinElute PCR purification kit (Qiagen) following the manufacturer’s instructions. DNA purified from the whole-cell extract sample was quantified using a NanoDrop 1000 Spectrophotometer. Genomic DNA ScreenTape was run on TapeStation (Agilent Technologies) to check the size distribution of fragmented chromatin. For the immunoprecipitated samples, extracted DNA was quantified using a Qubit dsDNA HS Assay Kit (Life Technologies). For preparing libraries for next-generation sequencing, procedures involving end repair, 3′ end adenylation, ligation with barcoded adaptors, and PCR amplification were performed with a TruSeq DNA Sample Preparation Kit or Ovation Ultralow system (NuGen). The amplified products within the range of 200–400 bp were size-selected by running a 2% (wt/vol) agarose gel with a BluePippin system (Sage Science). Gel-purified products were quantified by running D1000 ScreenTape on TapeStation (Agilent Technologies). Libraries were pooled and sequenced on the HiSeq 2000 platform for 100-bp single-end reads. Image analysis was performed in real time using HiSeq Control software (version 1.4.8) and Real-Time Analysis software (RTA; version 1.12.4.2) running on the instrument computer. Real-time base calling on the HiSeq instrument computer was performed with the RTA software. An Illumina CASAVA1.8 pipeline was used to generate the sequence data.

Reads from each sample were mapped to the mouse genome (mm10) using the Rsubread program (version 1.10.5) (59),with duplicate reads removed using SAMtools rmdup and bam files from replicate samples combined using SAMtools merge (66).

For Smchd1 peak identification, MACS2 peak calling was performed in WT NSCs (26). Data obtained from Smchd1^{MommeD1/MommeD1} NSCs were used as the negative control. The default settings were applied, with the exception that the P value cutoff was set as 5 × 10⁻⁴. Identified peaks were filtered based on the q value (false discovery rate). A total of 227 highly confident peaks (q < 0.1) were identified across the genome. The P value cutoff was relaxed to 5 × 10⁻³ for targeted peak identification within the Pcdh and Snrpn clusters.

For visualization of the ChIP-seq and MBD-seq results, signals between WT and Smchd1-null samples were normalized by read coverage calculated with the GenomicRanges Bioconductor package (67). Normalized signals were plotted on the y axis along the genomic location, with annotation obtained from Bioconductor package org.Mm.eg.db (version 3.0.0) (68), using the Gviz package (69). Coordinates for E14.5 whole-brain Ctcf ChIP peaks were obtained from the ENCODE database [University of California, Santa Cruz (UCSC) accession no. wgEncodeEM002595] (70). Coordinates were lifted from mm9 to mm10 mouse genome using the LiftOver tool from the UCSC genome browser (genome.ucsc.edu/cgi-bin/hgLiftOver) before being plotted along the genomic location. Coordinates for Ctcf ChIP peaks at the HS5-1 site were obtained from Monahan et al. (29) and were lifted from mm9 to mm10 mouse genome as described above (29).

For assessing Smchd1 peaks with respect to TSSs, the GREAT tool was applied (35). Coordinates for Smchd1 peaks were converted from mm10 to mm9 mouse genome and submitted to the program. The default “basal plus extension” setting was applied for associating Smchd1 binding sites with putative target genes. Distance between Smchd1 peaks and TSSs of putative target genes were calculated and categorized.

For assessing Smchd1 peaks with respect to annotated cis-regulatory elements, Smchd1 occupancy was compared with the H3K4me3-enriched region, RNA polymerase II binding sites, predicted enhancer sites, and Ctcf binding sites identified in E14.5 mouse brain (36). Coordinates for these cis-regulatory elements were downloaded from chromosome.sdsc.edu/mouse/download.html and converted from mm9 to mm10 mouse genome. Comparative analysis was computed with the GenomicRanges Bioconductor package (67).

For de novo motif analysis, DNA sequences corresponding to Smchd1 peaks were submitted to the MEME program (71). A total of 227 sequences (211.4 bp in length on average) were analyzed, with the minimum and maximum motif widths set as 6 bp and 35 bp, respectively. The top 10 retrieved motifs were submitted to TOMTOM for a similarity search within the JASPAR Vertebrates and UniPROBE Mouse motif database (71). Motifs 5 and 8 were matched to the Ctcf consensus sequence. Motif 4 was matched to the Rest/Nrsf consensus sequence.

For quantifying the H3K4me3, H3K27me3, and CpG methylation enrichment in Smchd1^{MommeD1/MommeD1} NSCs compared with the levels in WT NSCs, the relative log₂-fold change of the normalized number of reads was calculated using the featureCounts program (60). Reads aligned within the region spanning the Pcdh a1-a12 genes (chr18: 36925285–37025230) were compared between the genotypes for each ChIP-seq or MBD-seq experiment using the replicate samples. Similar analysis was performed for regions spanning the beta cluster (chr18: 37259998–37518652) and gamma cluster (chr18: 37656945–37824546).

Preparation of Whole-Cell Extracts.

Cultured NSCs were washed with cold PBS buffer and then lysed with KALB lysis buffer [150 mM NaCl, 50 mM Tris⋅HCl (pH 7.5), 1% (vol/vol) Triton X-100, 1 mM EDTA (pH 7.5)] supplemented with 1 mM sodium vanadate, 1 mM PMSF, and protease inhibitors (cOmplete protease inhibitor mixture tablets; Roche) on ice for 30 min. Insoluble material was removed by centrifugation at 15,000 × g at 4 °C for 5 min. Total protein concentration in the whole-cell extract was quantified using the BCA protein assay kit (Pierce) following manufacturer’s instructions.

Western Blot Analysis.

Proteins were resolved by 4–12% (wt/vol) SDS/PAGE (Invitrogen), transferred to PVDF membranes (Osmonics; GE), and blocked with 5% (wt/vol) skim milk powder in 0.1% (vol/vol) Tween 20-PBS for 1 h at room temperature. Membrane was incubated overnight with anti-Smchd1 antibody (1:2,000 diluted, A302-871A; Bethyl), anti-Ctcf antibody (1:2,000 diluted, 07-729; Millipore), and antitubulin antibody (1:2,000 diluted, sc-23948; Santa Cruz Biotechnology), respectively, at 4 °C, followed with HRP-conjugated secondary antibodies. Membrane was visualized using an ECL system (Luminata; Millipore) following the manufacturer’s instructions.

IP.

Whole-cell extract from one 10-cm² plate was incubated with 2.5 μg of anti-Smchd1 antibody (31865; Abcam) and anti-Ctcf antibody (SC15914; Santa Cruz Biotechnology) overnight at 4 °C for Smchd1 IP and Ctcf IP, respectively. A control IP procedure was set up with 2.5 μg of rabbit IgG (ab46540-1; Abcam). The antibody-enriched fraction was captured by incubation with 10 μL of Dynabeads Protein G (Life Technologies) at 4 °C for 1 h. The immunoprecipitated fraction was eluted with 2× SDS sample buffer and boiled at 95 °C for 5 min.

Recombinant Protein Expression.

Recombinant proteins were expressed and purified from BL21-CodonPlus expression competent Escherichia coli cells (Agilent). Cells were cultured in superbroth supplemented with 100 μg/mL ampicillin to an A₆₀₀ of ∼0.6–0.8 before expression was induced with 0.5 mM isopropyl β-d-1-thiogalactopyranoside overnight at 18 °C. Purification was performed essentially as described (72). Briefly, cells were lysed in lysis buffer [0.5 M NaCl, 20 mM Tris (pH 8), 20% (vol/vol) glycerol, 5 mM imidazole (pH 8), 0.5 mM Tris (2-carboxyethyl)phosphine], supplemented with 1 mM PMSF, by sonication. N-terminal 6-His–tagged proteins were purified by nickel-nitrilotriacetic acid (Ni-NTA) FastFlow resin (Qiagen) and eluted in the lysis buffer containing 250 mM imidazole (pH 8). The tag was cleaved by incubation with tobacco etch virus (TEV) protease for 1 h at room temperature. Cleaved protein was concentrated with a 30-kDa molecular mass cutoff concentrator (Millipore) at 4 °C by centrifugation at 3300 × g and then diluted with the lysis buffer. Subtractive Ni-NTA chromatography with the resin was performed to eliminate undigested protein and TEV protease, followed by a final Superdex-200 10/300 GL gel filtration (GE Healthcare) in 100 mM NaCl and 20 mM Hepes (pH 7.5). Fractions containing the Smchd1 hinge domain were pooled, concentrated, and then aliquoted and snap-frozen in a liquid nitrogen bath for storage at −80 °C.

EMSA.

Oligonucleotides (50 nM) were mixed with the recombinant Smchd1 hinge domain (WT or R1867G mutant) in a 0-, 10-, 50-, 250-, and 1,000-fold molar excess over the DNA in 1× PBS buffer in a total volume of 20 μL. Samples were incubated at room temperature for 30 min before addition of 5 μL of 50% (vol/vol) glycerol. The samples were then loaded onto a 0.5% (wt/vol) agarose gel in 1× Tris base, boric acid, EDTA (TBE) buffer and separated for 1.5 h at 4 V/cm at 4 °C. Gels were scanned on a Typhoon 9410 fluorescence scanner with 526-nm short-pass filter (GE Healthcare).

TSA.

For TSAs using a Corbett Real-Time PCR machine, proteins were diluted in 150 mM NaCl, 20 mM Tris (pH 8.0), and 1 mM DTT to a final concentration of ∼2–5 μM and assayed with 4–24 μM oligonucleotide in a total reaction volume of 25 μL. SYPRO Orange (Molecular Probes) at a 2× final concentration was used as a fluorescence probe and detected at 530 nm. The temperature was raised with a step of 1 °C per minute from 25 °C to 95 °C, and fluorescence readings were taken at each interval. For each well, sample fluorescence was plotted as a function of increasing temperature. The melting temperature corresponding to the midpoint for the protein unfolding transition was calculated by fitting the sigmoidal melt curve to the Boltzmann equation using GraphPad Prism, with R² values of >0.99. Data points after the fluorescence intensity maximum were excluded from fitting.

AUC.

All AUC experiments were conducted in a Beckman Coulter Model XL-I instrument set at a temperature of 20 °C. The protein and 6-FAM fluorescently labeled oligonucleotides were diluted to their respective concentrations in buffer [100 mM NaCl, 20 mM Hepes (pH 7.5)], loaded into double-sector quartz cells, and mounted in a Beckman Coulter eight-hole An-50 Ti rotor. Solvent density (1.0039 g/mL), viscosity [0.0103 centipoise (cp)], and partial specific volume of the protein (0.7399 mL/g) were computed using the amino acid composition and the program Sednterp (73). The partial specific volume ($\overline{v}$) of the protein/oligonucleotide complex was estimated to be 0.7180 mL/g, calculated with the equation

$$\overline{v}=\frac{{\text{M}}_1{\overline{v}}_1+{\text{M}}_2{\overline{v}}_2}{{\text{M}}_1+{\text{M}}_2},$$

using a partial specific volume of 0.55 mL/g and mass of 6.5 kDa for the oligonucleotides, and a partial specific volume of 0.7399 mL/g and dimeric mass of 50 kDa for the protein.

For sedimentation velocity experiments, 380 μL of sample and 400 μL of reference buffer were centrifuged at a rotor speed of 50,000 rpm. Data were collected at a wavelength of 488 nm, following the absorbance signal of the 6-FAM label, in continuous mode for 250 scans, using a time interval of 0 s and a step size of 0.003 cm without averaging. At this wavelength, we detect only the free oligonucleotide and Smchd1 hinge domain/oligonucleotide complex. Sedimentation velocity data at multiple time points were fitted to a continuous sedimentation coefficient [c(s)] model (74–76) using the program SEDFIT (available from www.analyticalultracentrifugation.com) to determine the ratio of bound and unbound oligonucleotides. The ratio of bound and unbound oligonucleotides was determined by the integration of peaks in the c(s) profile. In some cases, where the peaks in the c(s) distribution were not adequately baseline-resolved, the data were also fitted to a simple species analysis model using the program SEDFIT to determine the ratio of bound and unbound oligonucleotides. Using these ratios and the known initial concentrations of oligonucleotides and protein, the dissociation constant was calculated using the following formula:

$$K_{\mathrm{d}}=\left[\text{unbound}\hspace{0.17em}\text{oligonucleotide}\right]\left[\text{unbound}\hspace{0.17em}\text{protein}\hspace{0.17em}\text{dimer}\right]/\left[\text{protein}\hspace{0.17em}\text{dimer}/\text{oligonucleotide}\hspace{0.17em}\text{complex}\right].$$

This formula assumes that the stoichiometry of the interaction is one oligonucleotide to one Smchd1 hinge domain dimer, which we verify below.

The dissociation constants for both the WT and R1867G mutant of the Smchd1 hinge domain were determined using a variety of oligonucleotides. In each case, the concentration of the oligonucleotide was set at 1 μM and the protein concentration ranged from 1 to 40 μM. Initial concentrations of the oligonucleotides were determined by absorbance at 260 nm (ssDNA 15-mer poly-A, ε = 204,360 M⁻¹⋅cm⁻¹; ssDNA 15-mer poly-C, ε = 129,160 M⁻¹⋅cm⁻¹; ssDNA HS5-1b sense, ε = 204,860 M⁻¹⋅cm⁻¹; ssDNA HS5-1b antisense, ε = 219,060 M⁻¹⋅cm⁻¹; dsDNA HS5-1b, ε = 362,188 M⁻¹⋅cm⁻¹), and initial concentrations of the protein were determined by absorbance at 280 nm (Smchd1 and R1867G mutant, ε = 17,210 M⁻¹⋅cm⁻¹, and molecular mass of the monomer = 25.066 kDa). The K_d was determined for each sedimentation velocity run (at various protein and oligonucleotide concentrations) and then averaged; the error reported in Fig. 5F is the SEM for at least three separate measurements.

Control sedimentation velocity experiments were conducted to ensure that the fluorescent label (6-FAM) did not significantly contribute to binding of the oligonucleotide. Both the EMSAs (Fig. 5B) and AUC studies (Fig. S5C) for the ssDNA 15-mer poly-A 6-FAM–labeled oligonucleotide show poor binding to the Smchd1 proteins, demonstrating that the label does not mediate binding to the proteins. In addition, the presence of two 6-FAM labels, compared with one 6-FAM label, on several oligonucleotides did not affect the ratio of bound and unbound oligonucleotides in any way (Fig. S5A), and therefore the affinity.

The key advantage of using our AUC method to determine the K_d over traditional EMSAs is that we are able to assess the solution oligomeric state of the proteins, both free and when bound to the oligonucleotides. The Smchd1 hinge domain proteins were found to be dimeric at all concentrations tested (1–40 μM), irrespective of the presence of oligonucleotides. In fact, our simple method for determining the K_d assumes that there is one binding site per dimer; AUC is able to estimate the mass of the sedimenting complex by means of a continuous molar mass distribution [c(M)] analysis, using the sedimentation velocity data collected at high protein concentrations, where the predominant species (>90%) is the protein/oligonucleotide complex (Fig. S5B). Using a representative example, we found that the mass of the protein ssDNA HS5-1b sense unmethylated oligonucleotide complex is ∼53.4 kDa, consistent with one oligonucleotide bound per dimer. However, during the c(M) analysis, we found the frictional ratio for the complex was unstable, finding local minima in the fit between 1.28 and 1.40. This finding resulted in a peak in the c(M) that varied between 49.9 kDa and 56.9 kDa, but was still consistent with a stoichiometry of one oligonucleotide per dimer. Finally, we note that prior studies of the related Smc2/4 hinge domain heterodimer binding to fluorescently tagged oligonucleotides were also consistent with a stoichiometry of one protein dimer binding to one oligonucleotide (18).

To verify our method of determining the K_d constants detailed above, we used the traditional, although more time-consuming, method of sedimentation equilibrium with a representative oligonucleotide (ssDNA HS5-1b sense unmethylated oligonucleotide). For sedimentation equilibrium experiments, 110 μL of sample and 120 μL of reference buffer were centrifuged at rotor speeds of 15,000, 24,000, and 32,000 rpm. Data were collected at a single wavelength (488 nm) in step mode at a time interval of 0 s, a step size of 0.001 cm, and averaging 20 measurements until equilibrium was reached. These data best fitted the A + B ↔ AB hetero-oligomerization model (as implemented in SEDPHAT, www.analyticalultracentrifugation.com) compared with alternative models, giving good fits with largely random but small residuals to the individual experiments and a good global χ² value of 0.15. The K_d for the ssDNA HS5-1b sense unmethylated oligonucleotide in complex with WT Smchd1 hinge domain fitted to the A^(oligo) + B^(Smchd1) ↔ AB hetero-oligomerization model was 2.0 μM (68% confidence interval was 1.1–3.1 μM, determined using the F-statistic method). This result is in accordance with the K_d reported in Fig. 5F (2 ± 1 μM), demonstrating the utility of our relatively simple protocol for determining K_d constants for protein/oligonucleotide interactions.

Discussion

Here, we present the first, to our knowledge, high-resolution ChIP-seq analysis of Smchd1 binding, which we performed in male murine NSCs. Previous studies have predominantly used immunofluorescence and ChIP/quantitative PCR to describe colocalization of Smchd1 with H3K27me3, H3K9me3, and DNA methylation (2–5, 7, 9). Where ChIP-seq was reported (5), the data were analyzed in 150-kb nonoverlapping regions, possibly due to high background with their anti-SMCHD1 antibody, meaning that the data are of low resolution and would not identify sharp peaks such as we observed. Our high-resolution analysis of Smchd1 binding was enabled by using Smchd1-null samples as antibody controls; however, the background still may conceal more subtle features of Smchd1’s pattern of binding. Our results are consistent with previous reports but offer deeper insight, which, in concert with our biochemical and biophysical assays, allows us to propose models for Smchd1 binding to chromatin.

We find that Smchd1-bound regions are not restricted to promoters and that many are distant from the TSSs. A significant proportion of Smchd1 binding sites overlap with Ctcf binding sites, with many being putative cis-regulatory elements, including promoters, enhancers, and insulators, indicating that Smchd1 serves a broad array of roles in regulating gene expression. Specifically, Smchd1 binding at the HS5-1 enhancer site and promoters of the clustered Pcdh genes (particularly Pcdha1-a12), along with specific Ctcf-bound promoters and enhancers in the Snrpn cluster, correlate with alterations to other epigenetic marks and expression of genes within these clusters. In the absence of Smchd1, we observed both acquisition of active histone mark H3K4me3 and loss of repressive CpG methylation at individual promoters of these genes, concomitant with their elevated gene expression. However, Smchd1 may not be the sole factor in determining transcriptional outcome because gene regulation often involves multiple redundant pathways. This fact is exemplified by the finding that despite loss of Smchd1 binding at the Hox gene clusters, the polycomb group protein-mediated H3K27me3 and gene repression were maintained in Smchd1-null male NSCs.

We used the Pcdh cluster as a sensitive model to examine the genetic interaction between Smchd1 and Ctcf. We demonstrated a potential functional interaction between Smchd1 and Ctcf by showing their opposing action in regulating Pcdh gene expression. Ctcf is known to be involved in regulating gene expression via mediating long-range chromatin interactions and partners with an array of transcription factors, chromatin proteins, and RNA molecules (43, 45–47). It has been proposed that the Ctcf/Cohesin complex recruits selected Pcdh alpha promoters to the HS5-1 site, and thus creates an active transcriptional hub for maximal expression (30) (Fig. 6, Left). In accordance with previous studies (30, 48), we observed stable CpG methylation at Pcdha1-a12 promoters in WT NSCs that could potentially hinder Ctcf binding. On the other hand, additional Ctcf binding was observed at the hypomethylated promoters in Smchd1-null NSCs. Together with the discovery of Smchd1’s localization at promoter regions and the HS5-1 site, these results have led us to hypothesize that Smchd1 may facilitate a repressive domain that potentially shields promoters of Pcdh genes from Ctcf binding (Fig. 6, Middle). This model is further supported by our in vitro DNA binding analyses using EMSA together with complementary biophysical assays. However, it is unclear whether Smchd1 and Ctcf act in a competitive manner or whether they could simultaneously co-occupy the HS5-1 site in a cooperative way, where Smchd1 restrains the inactive promoters from the active transcriptional hub (Fig. 6, Right). We propose that Ctcf selectively binds the unmethylated promoters, whereas Smchd1 engages with the methylated promoters to maintain their DNA methylation and repressed status; this model may also hold true at the Snrpn-imprinted cluster, given that Smchd1 binding is enhanced by DNA methylation for at least one site in this cluster. Our data support the idea that without Smchd1, Pcdh promoters are unleashed from a repressive chromatin environment, with pursuant engagement of Ctcf/Cohesin resulting in activation, at least for many of the Pcdh alpha genes.

Fig. 6.

Schematic diagrams for coordinated regulation of Pcdh alpha genes by Smchd1 and Ctcf through competitive (Left and Middle) or cooperative (Right) models. (Left) Active transcriptional hub model is based on a previous study by Guo et al. (30), where Ctcf/Cohesin-mediated long-range chromatin interactions bring active Pcdh alpha genes into close proximity to the HS5-1 enhancer site. (Middle) Smchd1 binding may facilitate or maintain a repressive domain antagonizing Ctcf binding. (Right) Alternatively, Smchd1 and Ctcf could co-occupy the HS5-1 site, and Smchd1 might restrain the inactive promoter from the active transcriptional hub.

In addition to overlap with Ctcf sites, we have identified a subset of Smchd1 binding sites that overlap with Rest/Nrsf occupancy. Rest/Nrsf is a transcription repressor that suppresses neural-specific gene expression in nonneuronal tissues (49–51). It would be interesting to examine if Smchd1 cooperates with Rest/Nrsf in this process in the future. Beyond the clustered Pcdh genes, the functional significance of Smchd1 binding at other genomic loci, for example, at the Snrpn locus, is the subject of ongoing investigation.

To characterize how Smchd1 binds to chromatin, we used a variety of biochemical and biophysical analyses. These analyses have provided the basis for characterizing the exact DNA binding mode of the hinge domain of Smchd1. Our AUC data have revealed several previously unrecognized features of Smchd1: first, the Smchd1 hinge domain dimerizes, similar to the hinge domains from other SMC proteins (16, 17, 20); second, Smchd1 might directly associate with DNA via its hinge domain; third, Smchd1 has the potential to bind methylated DNA, consistent with its predicted function in maintaining DNA methylation (2, 4, 7); and, finally, Smchd1 has the potential to bind RNA. These data are consistent with our model of Smchd1 and Ctcf binding at the Pcdh alpha cluster, and they raise the intriguing possibility that like Ctcf (43), Smchd1 may also partner with RNA molecules for part of its epigenetic function. Given Smchd1’s unique protein domain arrangement, with a C-terminal SMC hinge domain that directs chromatin interactions and an N-terminal putative ATPase domain, we are keen to investigate the possibility that Smchd1 exerts active manipulation at the chromatin level, and whether this function somehow requires RNA interaction.

In conclusion, our results reveal a potential involvement of Smchd1 in long-range chromatin interaction-mediated epigenetic regulation. We have established a functional link between Smchd1 and Ctcf in regulating expression of the Pcdh genes. The coincidence of Smchd1’s genome-wide occupancy with a subset of Ctcf binding sites raises the intriguing possibility that Ctcf and its plethora of interacting factors may have some bearing on how Smchd1 regulates transcription. Interestingly, CTCF may also have the opposite effect of SMCHD1 in facioscapulohumeral muscular dystrophy (52). Therefore, further characterization of Smchd1’s DNA binding sites and in vivo RNA binding partners in other cellular contexts will provide deeper insights into the molecular mechanism of Smchd1-mediated gene regulation and the underlying implications for development and disease.

Methods

All experimental animals were treated in accordance with the Australian Government National Health and Medical Research Council guidelines under approval from the Animal Ethics Committees of the Walter and Eliza Hall Institute (WEHI AEC 2011.027) and the Queensland Institute of Medical Research (QIMR AEC A0812-610M). Additional details are provided in SI Methods.

Derivation and Culture of NSCs.

Brains from E14.5 embryos derived from C57BL/6 Smchd1^MommeD1/+ males mated with FVB/N Smchd1^MommeD1/+ females were dissected out in Leibovitz’s L-15 Medium (Gibco) using NSCs as described previously, and as detailed in SI Methods (53).

Transcriptome Sequencing and Analysis.

RNA and DNA were extracted from WT and Smchd1^{MommeD1/MommeD1} male NSCs using an AllPrep DNA/RNA Mini Kit (Qiagen) according to the manufacturer’s instructions and were quantified using a NanoDrop 1000 Spectrophotometer (Thermo Scientific). RNA integrity was assessed with an Agilent Bioanalyzer 2100 (Agilent Technologies) according to the manufacturer’s instructions. Sequencing libraries were prepared with a TruSeq Stranded Total RNA Sample Preparation Kit (Illumina) according to the manufacturer’s instructions. Sequencing and analysis were as described in SI Methods.

MBD-Seq.

Genomic DNA was extracted from WT and Smchd1^{MommeD1/MommeD1} male NSCs. After purification and fragmentation, methylated DNA was isolated using the MethylMiner Methylated DNA Enrichment Kit (Life Technologies) according to the manufacturer’s instructions. The DNA was eluted from MBD-coupled beads in two salt concentration cuts: first, a 600 mM NaCl cut to remove poorly methylated DNA, followed by a 2 M NaCl cut to elute highly methylated DNA; the latter was used for the preparation of libraries for next-generation sequencing Details are provided in SI Methods.

ChIP-Seq and Analysis.

ChIP was performed with the fast ChIP protocol (54) with modifications as detailed in SI Methods. Antibodies used were as follows: Smchd1 (31865; Abcam), H3K4me3 (07-473; Millipore), H3K27me3 (07-449; Millipore), and Ctcf (07-729; Millipore). Library preparation and sequencing were performed using standard protocols (SI Methods). Analysis details are provided in SI Methods.

Cloning, Expression, and Purification of Recombinant Protein.

cDNA encoding amino acids 1683–1899 of Smchd1 was PCR-amplified from a full-length cDNA clone and cloned into a pPROEX HTb vector (Life Technologies) for expressing 6-His–tagged recombinant protein. The R1867G mutation was introduced by oligonucleotide-directed PCR mutagenesis. Details of the expression and purification procedure are provided in SI Methods.

EMSA.

The EMSA was performed with 6-FAM fluorescence-labeled and HPLC-purified oligonucleotides (Integrated DNA Technologies) dissolved in 10 mM Tris⋅HCl (pH 8.5). The dsDNA was annealed by mixing equal volumes of 100 μM ssDNA (strands 1 and 2), incubated at 95 °C for 5 min, and then gradually cooled down to room temperature. The EMSA was performed using the method described by Griese et al. (18) and is detailed in SI Methods.

TSA.

The TSA was performed using the method described by Murphy et al. (42), with modifications as detailed in SI Methods.

AUC.

AUC experiments were conducted using a Beckman Coulter model XL-I instrument at 20 °C. Protein and 6-FAM fluorescently labeled oligonucleotides were mixed in 100 mM NaCl and 20 mM Hepes (pH 7.5), loaded into double-sector quartz cells, and mounted in a Beckman Coulter eight-hole An-50 Ti rotor. Details of the experiments and analyses are included in SI Methods, and additional details are shown in Figs. S5 and S6.