The non-canonical SMC protein SmcHD1 antagonises TAD formation on the inactive X chromosome

The inactive X chromosome (Xi) in female mammals adopts an atypical higher-order chromatin structure, manifested as a global loss of local topologically associated domains (TADs), and formation of two mega-domains. In this study we demonstrate that the non-canonical SMC family protein, SmcHD1, which is important for gene silencing on Xi, contributes to this unique chromosome architecture. Specifically, allelic mapping of the transcriptome and epigenome in SmcHD1 null cells revealed the appearance of sub-megabase domains defined by gene activation, CpG hypermethylation and depletion of Polycomb-mediated H3K27me3. These domains, which correlate with sites of SmcHD1 enrichment on Xi in wild-type cells, additionally adopt features of active X chromosome higher-order chromosome architecture, including partial restoration of TAD boundaries. Xi chromosome architecture changes also occurred in an acute SmcHD1 knockout model, but in this case, independent of Xi gene de-repression. We conclude that SmcHD1 is a key factor in antagonising TAD formation on Xi.


Introduction
X chromosome inactivation is the mechanism that evolved in mammals to equalise levels of X-linked gene expression in XX females relative to XY males. Cells of early female embryos selectively inactivate a single X chromosome, usually at random, resulting in the formation of a stable heterochromatic structure, the Barr body. The inactive X chromosome (Xi), once established, is highly stable, and is maintained in somatic cells throughout the lifetime of the animal 1,2 . The X inactivation process is triggered by the non-coding RNA Xist, which localises to the Xi territory to induce chromosome-wide gene silencing [3][4][5][6] .
Chromatin features that distinguish Xi and the active X chromosome (Xa) include specific histone post-translational modifications, variant histones, and CpG DNA methylation (reviewed in 2 ). Additionally, Xi acquires a characteristic higher-order chromosome structure. Specifically, A-type chromatin compartments, corresponding to gene-rich regions which normally replicate in early S-phase, switch to replication in mid-or -late S-phase (reviewed in 7 ). Additionally, topologically associated domains (TADs), sub-megabase scale domains which are formed by the activity of cohesin, restricted at boundaries by oppositely oriented binding sites for the insulator protein CTCF [8][9][10][11][12][13] , are in large part absent on Xi, being replaced instead by two large mega-domains that are separated by a hinge that encompasses the DXZ4 repeat sequence [14][15][16][17][18] . The basis for this unique TAD structure is not well understood, but is thought to depend, at least in part, on ongoing expression of Xist RNA 17 .
Barr body formation is a multistep process. Thus, Xist RNA recruits specific chromatin modifiers, including the SPEN-NCoR-HDAC3 complex [19][20][21][22] , required for histone deacetylation 22 , and the PRC1 and PRC2 Polycomb complexes, required for deposition of H2A lysine 119 ubiquitylation (H2AK119u1) and H3 lysine 27 methylation (H3K27me3) respectively [23][24][25][26][27] . The lamin B receptor 22,28 , and m6A RNA modification complex 19,29 , have also been implicated in establishment of chromosome-wide gene silencing. Other factors are recruited to Xi at later stages. Examples include the variant histone macroH2A 30 , and the non-canonical SMC protein SmcHD1 31 . The role of these factors remains to be defined, although is likely to be linked to the long-term stability of the inactive state.
SmcHD1 is classified as an SMC protein by virtue of an SMC hinge domain at the Cterminal end, but differs from canonical SMC complexes in having a functional GHKL-ATPase domain rather that a Walker A/B type ATPase domain 32 . Biochemical and biophysical studies indicate that SmcHD1 homodimerises via the hinge and GHKL domains to form a complex that is reminiscent of bacterial SMC proteins, both in form and scale 33 , albeit forming a functional homodimer rather than a trimeric complex. SmcHD1 performs an important role in silencing on Xi, and at selected mono-allelically expressed autosomal loci 31,32,34,35 . Whilst it is known that a proportion of Xi genes are activated in SmcHD1 mutant embryos 34,35 , the molecular mechanism is not well understood. Notably, although SmcHD1 is required for DNA methylation at CpG island (CGI) promoters of many Xi genes, loss of CGI methylation does not appear to account for the observed gene activation 34 . An alternative hypothesis is that SmcHD1-mediated compaction of Xi, inferred by microscopy based analyses in human cell lines 36 , imposes gene repression. Given the important role of SMC family proteins in genome topology, we set out to investigate the role of SmcHD1 in the higher-order architecture of Xi. Thus, we performed high-resolution analysis of Xi transcription, epigenetic features, and higher-order chromatin features in SmcHD1 null cell lines. Our findings demonstrate that SmcHD1 on Xi plays a role in transcriptional and epigenetic regulation of sub-megabase chromosome domains, and in antagonising TAD formation.

MEFs.
In order to gain insight into the role of SmcHD1 in X inactivation we set out to analyse epigenomic and long-range chromatin features of Xi at high resolution. Thus, we derived XX MEF lines from Mus musculus domesticus (domesticus) x Mus musculus castaneus (castaneus) female embryos, that were either wild-type (WT) or SmcHD1 null. Xi was of castaneus origin in both cases ( Figure S1a,b). The high frequency of SNPs between domesticus and castaneus genomes allows assignment of high-throughput sequencing reads to either maternal or paternal genomes. The breeding strategy enabled us to obtain cell lines from F2 embryos in which the entire X chromosome was either of domesticus or castaneus origin. Autosomes on the other hand were mosaic as a result of recombination in the F1 generation. X inactivation in the interspecific embryos is random, so stable MEF lines were sub-cloned to obtain WT and SmcHD1 null lines. Xist RNA FISH and karyotype analysis confirmed the presence of Xi and Xa chromosome(s) (Figure S1c-e).
Initially, we performed allelic ChIP-seq analysis of SmcHD1 to define binding sites on Xi. As shown in Figure 1a, SmcHD1 is highly enriched over domains that correlate with gene rich regions across the length of the chromosome. We also observed SmcHD1 enrichment over specific regions on autosomes, including known SmcHD1 target loci such as the protocadherin locus on chromosome 18 and the PWS/AS imprinted gene cluster on chromosome 7 ( Figure S2a,b). SmcHD1 peaks in the latter region accord with a previously published dataset analysing SmcHD1 occupancy in XY neuronal stem cells 37 .
We went on to analyse Xi transcription in SmcHD1 null compared to WT cells using allelic chromatin RNAseq (ChrRNA-seq), which enriches for nascent unprocessed mRNAs, thereby maximising the number of informative SNPs due to inclusion of intron sequences.   The results are summarised in Figure 1b-e. In WT cells, Xi expression was detected for a small number of genes (28/404), largely corresponding to genes previously reported to escape X inactivation ( Figure S2c). However, in SmcHD1 null cells, Xi expression was seen for the majority of Xi genes (254/404) (Figure 1b-e). For 67 SmcHD1-dependent genes, Xi expression was in a similar range to that found on Xa, with a further 163 genes showing partial or low-level activation (Figure 1b-e, Figure S2c). We observed extensive overlap with 64 SmcHD1 dependent genes identified in a prior study in which we performed nonallelic microarray analysis of SmcHD1 null embryos 34 ( Figure S2d). The larger number of Xi expressed genes observed in this study can be attributed to the increased sensitivity afforded by allelic ChrRNA-seq.
Analysis of the association of SmcHD1-dependent genes with genomic sequence features in the immediate chromosomal environment identified a correlation with gene-rich regions, and the location of SINE repeats, as reported previously 34 Figure S3a-c). We then performed ChrRNA-seq as described above. Interestingly, and in contrast to embryonic SmcHD1 loss of function, no activation of Xi genes was observed ( Figure S3d). This result suggests that SmcHD1 is required to reinforce gene silencing during a specific window in development, and is then dispensable, presumably reflecting compensation through other maintenance pathways.

Unique features of the Xi epigenome in SmcHD1 null MEFs
Previous work established that SmcHD1 is important for DNA methylation of Xi CGI promoters 31 , but its role in chromosome-wide intergenic and intragenic DNA methylation on Xi has not been investigated. To address this we used whole genome bisulfite analysis (WGBS) to determine the methylome of Xa and Xi at single nucleotide resolution in WT and SmcHD1 null MEFs. A summary of data illustrating overall methylation density in 100kb windows across the entire genome is shown in Figure 2a. CpG methylation across autosomes was generally in the range of 60-80%, but was significantly lower on the X chromosome. We noted that gene-poor regions are CpG hypomethylated across all chromosomes. An example, chromosome 7 is illustrated in Figure S4a. A similar pattern of hypomethylation is apparent in available MEF WGBS 38 ( Figure S4b). Accordingly, we find that total CpG methylation levels in MEFs, as determined by HPLC analysis, are moderately reduced relative to ES cells and adult tissue ( Figure S4c). The molecular basis for this pattern of CpG hypomethylation is currently unknown.
We went on to compare allelic methylation levels on Xa and Xi chromosomes ( Figure   2b). In WT MEFs, Xa methylation patterns were similar to autosomes (56% compared with 66% CpG methylation), but Xi was extensively CpG hypomethylated across the entire chromosome (19% CpG methylation). CpG hypomethylation of gene-poor regions, as described above, likely contributes to the observed pattern on Xi. However, CpG hypomethylation was also evident in gene-rich regions, presumably linked to X inactivation status. This observation is consistent with prior studies showing reduced levels of CpG hypomethylation on Xi relative to Xa 39,40 . Xi CGIs in WT MEFs were, as expected, highly methylated ( Figure 2c). Additionally, we observed high levels of CpG methylation over bodies of genes that escape X inactivation ( Figure S4d).
In SmcHD1 null MEFs, Xi CGIs were in most cases hypomethylated (Figure 2c), as previously reported 31 . Conversely, we observed several domains with relatively high CpG methylation ( Figure 2b). These domains correspond in most cases with Xi genes that are   (Figure 2d,e). Individual genes that normally escape X inactivation were also CpG hypermethylated on Xi, similar to WT MEFs ( Figure S4d).
The histone modification H3K27me3 catalysed by the major Polycomb complex PRC2, is highly enriched on Xi as determined by immunostaining 23,24 , and high-resolution ChIP-seq analysis 41 . Previously we noted that this feature is not grossly affected by SmcHD1 loss of function, as determined by immunostaining of interphase nuclei in XX embryos 32 .
Similarly, H3K27me3 enrichment on Xi was readily detected by immunostaining in the WT and SmcHD1 null MEF lines described herein ( Figure S3b). However, allelic ChIP-seq analysis revealed domains in which Xi H3K27me3 is markedly depleted in SmcHD1 null cells

A modified higher-order chromosome architecture on Xi in SmcHD1 null cells
In light of evidence that SmcHD1 influences Xi chromatin at the level of sub-megabase domains, we went on to directly analyse parameters of long-range chromatin architecture.
Mammalian chromosomes are comprised of distinct gene-rich and gene-poor compartments with sizes ranging from sub-megabase to several megabases in length. These regions are replicated co-ordinately, either during early-or mid/late-S phase respectively. Replication  timing domains are broadly synonymous with A-and B-type chromatin, which in studies on long-range chromosome topology, have been shown to self-associate 42 . Additionally, Btype domains overlap extensively with Lamin associated domains (LADs) which localise to the nuclear periphery 43 . Xi is unusual in that both gene-rich and gene-poor compartments replicate synchronously in mid/late S-phase 7 .
A previous analysis of a human XX somatic cell line in which SMCHD1 was depleted using siRNA, revealed an aberrant replication timing pattern for Xi, with the appearance of early replicating domains as determined using a cytogenetic assay 36 . With this observation in mind we set out to determine allelic temporal replication patterns in our WT and SmcHD1 null MEFs using RepliSeq, a high-resolution sequencing based approach 44 ( Figure S6a Analysis of data for autosomes and for Xa indicated that SmcHD1 does not impact on global TAD structure. An example, chromosome 7, is illustrated in Figure S7a      c. Difference between normalised interaction counts of WT and SmcHD1 null MEFs.
Heatmap for Xi presents additionally quantification of differences in interaction counts between WT and SmcHD1 null MEFs for distinct parts of the Chr X interaction matrix d. Local chromatin conformation of the distal 20 Mb of Xa and Xi for WT and SmcHD1 null MEFs. Heatmaps were generated as above, but interactions were binned into 100 kb bins.

SmcHD1 depletes CTCF and cohesin at TAD boundaries on Xi
Recent studies have established that TAD boundaries result from the insulator protein CTCF restraining processive activity of the cohesin complex [11][12][13] . Accordingly, CTCF and cohesin occupancy is reduced at many sites on Xi compared with Xa 17,46,47    To assess whether changes in CTCF/Rad21 occupancy depend on Xi gene activation, we analysed the acute SmcHD1 null MEF line in which there is no detectable Xi gene de-repression. Interestingly, restoration of CTCF/Rad21 occupancy on Xi was also evident in this cell line ( Figure S8a,b), although Rad21 occupancy was of a lesser magnitude compared to MEFs derived from SmcHD1 null embryos (Figure 6b,d,e). This result suggests SmcHD1 directly affects TAD formation on Xi, independent of its role in maintaining Xi gene silencing.

Discussion
Allelic ChIP-seq analysis of SmcHD1 on Xi shows strong enrichment over gene-rich domains. This pattern mirrors the localisation of Xist RNA and Xist-dependent histone modifications 48,49 , and is therefore consistent with previous studies that reported a requirement for ongoing Xist expression to maintain SmcHD1 enrichment on Xi 36 . Using allelic RNA-seq we found that a large proportion of Xi genes within SmcHD1 enriched regions are de-repressed in SmcHD1 null MEFs, albeit, in most cases expression does not reach the level seen on Xa. At present we cannot discriminate between the possibility that Xi gene activation in SmcHD1 null cells occurs in a graded fashion within individual cells, or that low level activation reflects probabilistic expression in a subset of cells that varies for different loci.
In prior work we found that SmcHD1 enrichment on Xi is a late step in the X inactivation cascade in differentiating XX ESCs 31 , suggesting a role in maintenance rather than establishment of X inactivation. Paradoxically, in this study we found that there is no activation of Xi genes following acute knockout of SmcHD1 in MEFs. In light of this we speculate that SmcHD1 is required for Xi to transition from the initial repressed state to longterm silencing, but that downstream silencing pathways, for example CGI DNA methylation, maintain the inactive state even in the absence of SmcHD1. Although we cannot at present define how SmcHD1 mediates transitional Xi gene silencing, its role in establishing higher-order chromosome folding, discussed below, is one possible mechanism.
Analysis of the Xi methylome at single nucleotide resolution confirms prior studies demonstrating Xi hypomethylation 39,40 . We observed widespread and extensive hypomethylation, encompassing both gene-rich and gene-poor compartments. As noted, we cannot rule out that hypomethylation of gene-poor regions on Xi is linked to genome-wide effects in MEFs. Hypomethylation of gene-rich regions on Xi on the other hand, is presumably linked to X inactivation status. Mechanistically, this could be attributable to reduced binding of the de novo DNA methyltransferases Dnmt3a/b, which recognise the histone modification H3K36me3 in gene bodies of active genes 50,51 . Consistent with this idea, hypermethylated regions on Xi correspond to genes that escape X inactivation and in SmcHD1 null MEFs, hypermethylation occurs within regions encompassing activated Xi genes. However, Xi CpG hypermethylation in SmcHD1 null MEFs is not restricted to gene bodies, where H3K36me3 is normally enriched, but rather occurs across domains that include upstream and downstream sequences, and in some cases more than one gene.
Reciprocal depletion of PRC2-mediated H3K27me3 further reinforces that deregulation of the Xi epigenome in SmcHD1 null cells occurs at the level of sub-megabase domains.
These observations suggest alternative mechanisms, for example increased accessibility to nucleosome remodelling complexes, are involved in establishing the distinct epigenomic features observed on Xi in SmcHD1 null MEFs.
Further support for SmcHD1 functioning on Xi at the level of domain organisation comes from direct analysis of replication-timing domains and TADs. In both cases we observe a shift towards Xa-specific organisation across Xi, most obviously associated with sites where Xi gene activation and modified epigenomic features occur. The mechanism for replication-timing changes is not known. One possibility is that SmcHD1 affects recruitment of Rif1, required to direct PP1 to reverse Cdc7-mediated phosphorylation of the MCM complex 52 . Thus, at chromosomal regions that are normally early replicating, SmcHD1 may facilitate Rif1 recruitment, for example through affecting chromosome architecture.
SmcHD1 has been reported to disrupt Xi replication-timing independent of gene activation following SMCHD1 knockdown in a human XX cell line 36 , implying that transcriptional activation is not the cause of shifted replication-timing patterns.
The unique organisation of TADs on Xi has been linked to reduced binding of CTCF at TAD boundaries and/or reduced recruitment of cohesin complexes 17,18 . Accordingly, we observe restoration of both CTCF binding and cohesin (Rad21) accumulation in SmcHD1 null MEFs, and sites of restoration correlate well with the reappearance of Xa TAD structure.
These findings accord with prior analysis of SmcHD1 at an autosomal target, the protocadherin gene cluster in neural stem cells, which suggested that SmcHD1 and CTCF have opposing roles in transcriptional regulation 37 . The antagonistic effects of SmcHD1 on TAD boundaries may be analogous to the role of condensin in TAD dissolution during mitosis 53 .
We observed that whilst restoration of CTCF binding also occurs in MEFs following acute deletion of SmcHD1 (in which there is no Xi gene de-repression), restoration of Rad21 on Xi was of a lower magnitude. Prior work has suggested a link between cohesin loading and active transcription 54 , and this may account for the reduced level of restoration relative to the chronic SmcHD1 null model. Regardless, this result indicates that changes in the long-range architecture of Xi in SmcHD1 null cells is not a consequence of gene derepression. We note that prior studies have reported that depletion of Xist RNA in somatic cells leads to partial restoration of Xa chromosome architecture/TAD structure in the absence of changes in Xi gene repression 17,55 . Given that SmcHD1 recruitment to Xi is dependent on ongoing Xist expression 36 , we suggest that these effects may be attributable in part, or entirely, to loss of SmcHD1 on Xi.
In summary, our findings illustrate that SmcHD1 functions in X inactivation at the level of chromatin domain organisation. In future studies it will be important to determine the molecular interactions that underpin SmcHD1 function in antagonising TAD formation both on Xi and potentially at other SmcHD1 target loci.

Derivation and culture of mouse embryonic fibroblasts (MEFs)
Animal studies were carried out under the United Kingdom Home Office ASPA project, licence numbers 30/2800 (until 2015) and 30/3326 (from 2015 to present).
Interspecific crosses between SmcHD1 mutation carrier strain (FVB) and castaneus strain were designed to prevent meiotic recombination between X cast and X FVB chromosomes ( Fig.1)  neighbouring exons were cloned into plasmid pX459 as described 57 . SmcHD1 wt1A2 cell line that carries WT SmcHD1 (two FVB alleles and one castaneus allele) was used for mutagenesis. Fibroblasts were plated on 90mm Petri dishes a day before transfection. A few hours before transfection growth medium was replaced with antibiotic-free medium. Cells were transfected with a pool of two sgRNAs, 3g each, using Lipofectamine 3000 (Life Technologies) according to the manufacturer's instructions. DNA:Lipofectamine ratio was 1:3. Cells were trypsinised 18hrs after lipofection and plated on 145mm petri dishes at densities 1/10; 1/3 and the rest. Puromycin selection at final concentration 4μg/ml was applied next day and maintained for 72hrs, after which cells were grown in EC10 medium for 10 days until colonies were ready to be picked. Individual well-spaced colonies were scraped from the dishes and transferred into 48-well plate without trypsinisation. Initial screening of knockout clones was by genomic PCR across intronic region (SmcHD1_TNK208+ SmcHD1_TNK291). Selected clones were subsequently analysed by sequencing of cloned PCR products from regions around sgRNAs and across introns when appropriate. Candidate KO clones D4 and A3.3 were subsequently further characterised by immunofluorescence with SmcHD1, H3K27me3 and H2Aub1 antibodies, and by Western blot of nuclear extracts with SmcHD1 antibody.

Metaphase spreads
Cells were plated on 140mm Petri dishes two days before metaphase collection. Semiconfluent cultures with actively dividing cells were fed with fresh medium supplemented with

M-FISH
Metaphase spreads were prepared as described above, omitting the addition of ethidium bromide. The cells were hybridised with the 21XMouse MFISH kit (Zeiss Metasystem), following the manufacturer instructions. The slides were mounted in DAPI/Vectashield mounting medium (Oncor), under a glass coverslip, and analysed with an Olympus BX60 microscope for epifluorescence equipped with a Sensys CCD camera (Photometrics, USA).
Images were collected using Genus Cytovision software (Leica). A minimum of twenty-five cells were analysed for each cell line.

Chromosome paint (DNA FISH) on metaphase spreads
Slides with freshly fixed metaphase spreads were dehydrated through ethanol series (2x70%, 2x90% 2 min each followed by 100% for 5 min), and incubated in an oven at 65 0 C for 1hr. Slides were cooled down on a bench for a few minutes and denatured in 70% (v/v) Formamide/2xSSC at 65 0 C for 90sec. Slides were quenched in ice-cold 70% ethanol for 4 min and then dehydraded again through the ethanol series (the same as above), and dried in the vacuum dessicator for 5 min at RT. A 1:1 mix of directly labelled chromosome 8 (Cy-3, Cambio Ltd) and X (FITC) paints was denatured at 65 0 C for 10min, spun down and incubated at 37 0 C for 40 to 60min. Probe was incubated with the denatured metaphases overnight at 37 0 C. Next day the slides were washed twice with a solution of 1xSSC/50% formamide followed by two washes with 1xSSC and three washes with 4xSSC/0.05% Tween-20 in a water bath at 45ºC, 5 min each. Slides were mounted in Vectashield containing 4,6-diamidino-2-phenylindole (DAPI) (Vector laboratories) and sealed with nail varnish.

RNA-FISH
Cells were plated on Superfrost Plus gelatinised slides (VWR) and grown at least overnight.
Slides were rinsed briefly in PBS and fixed in 4% formaldehyde/PBS for 10 mins on ice, followed by two washes in 70% ethanol. Slides were either stored in 70% ethanol at 4 0 C until use or dehydrated (80%, 95%, 100% ethanol, 3 min each, RT) and air dried immediately before hybridisation with Xist probe. Xist probe was generated from an 18 kb cloned cDNA spanning the whole Xist transcript using a nick translation kit (Abbott Molecular) as previously described 19 . Directly labelled probe (1.5μL) was co-precipitated with 10μg salmon sperm DNA, 1/10 volume 3M sodium acetate (pH 5.2) and 3 volumes of 100% ethanol. After washing in 75% ethanol, the pellet was dried, resuspended in 6μL deionised formamide and denatured at 75ºC for 7 min before quenching on ice. Probe was diluted in 6μL 2x hybridisation buffer (5x SSC, 12.5% dextran sulfate, 2.5mg/mL BSA (NEB)), added to the denatured slides and incubated overnight at 37ºC in a humid chamber. After incubation, slides were washed three times with a solution of 2xSSC/50% formamide followed by three washes with 2xSSC in a water bath at 42 ºC. Slides were mounted with Vectashield with DAPI and sealed with nail varnish.

Immunofluorescence
Cells were plated on Superfrost Plus gelatinised slides (VWR) at least a day before the experiment. On the day of the experiment, cells were washed with PBS and then fixed with 2% formaldehyde in PBS for 15 mins at RT, followed by 5 mins of permeabilisation in 0.4% Triton X-100. Cells were rinsed with PBS three times, 2 min each and pre-blocked with a 0.2% w/v PBS-based solution of fish gelatine (Sigma) 3 times, 5 min each. Primary antibody dilutions were prepared in fish gelatin solution with 5% normal goat serum. Primary antibody dilutions are listed below. Cells were incubated with primary antibodies for 2 hours in a humid chamber at room temperature, then washed three times in fish gelatin solution to remove non-bound and non-specifically bound antibodies. Secondary antibodies were diluted in fish gelatin solution and incubated with cells for 1hr at RT in a humid chamber.
After incubation, slides were washed twice with fish gelatin and once with PBS before mounting with Vectashield mounting medium with DAPI. Excess mounting medium was removed and the coverslips were sealed using nail varnish.

Microscopy
Z stack images were acquired on a Zeiss AX10 microscope equipped with AxioCam MRm charge-coupled device camera using AxioVision software (Carl Zeiss International, UK).
Best exposure time for each field and channel was manually determined and kept fixed among experiments. Further image editing and refinement was achieved through Fiji/ImageJ.

RNA extraction
Cells grown on T25 flasks (Nunc) were washed twice in PBS and lysed directly with 1ml of Trizol reagent (Life Technologies). Samples were incubated for 5 min at 25ºC, cellular lysates were collected and transferred into 1.5ml RNase-free Eppendorf tubes. 0.2 volume of chloroform was added to each sample and mixed by vigorous shaking for 15 s. Samples were incubated for 2 min at 25ºC and centrifuged at 12,000g for 5 min at 4 ºC. The upper aqueous phase was transferred into a clean tube and an equal volume of isopropanol was added. Samples were incubated for 10 min, and RNA was pelleted at 12,000g for 10 min at 4ºC. The pellets were washed once in 1ml of 75% ethanol and then air-dried for 5-10 min and resuspended in 50-100μL RNase-free water. Contaminating DNA was removed using the Ambion DNA-free DNase Treatment kit (Life Technologies) according to the manufacturer's instructions. cDNA was generated with pd(N)6 random hexamers (GE Healthcare) using SuperScript III reverse transcriptase (Life Sciences) according to the manufacturer's instructions.

Genomic DNA extraction
Cells from a confluent 90-140mm petri dish were harvested and resuspended in 5-10ml lysis buffer (10 mM NaCl, 10 mM Tris-HCl pH 7.5, 10 mM EDTA-NaOH pH 8.0, 0.5% Sodium lauroyl sarcosinate) with proteinase K added to a final concentration of 100μg/ml. Samples were incubated overnight at 55ºC, and then genomic DNA was phenol/chloroform extracted and purified using 15 ml MaXtract High Density Tubes (Qiagen). Genomic DNA was precipitated with 1/25 volume of 5M NaCl and 2.5 volume of ice-cold 100% ethanol. High molecular weight genomic DNA was spooled and transferred to a new Eppendorf tube containing 1ml of 70% ethanol. DNA was pelleted, air dried and resuspended in 300-400μl 10 mM Tris pH 8.5. DNA concentration was measured by Nanodrop.

Nuclear extraction
Nuclear extracts for Western blot analysis of CRISPR/Cas9-mediated SmcHD1 mutant cell lines were prepared essentially according to a method described previously 58

Repli-seq
Repli-seq was performed as described 60 . Briefly, asynchronously proliferating cells were flow-sorted based on their DNA content (Hoechst 33342) into G1 and S phase fractions

HiC
Hi-C was performed as described 61 . For each Hi-C library 25 x 10^6 cells were cultured in EC10 media. The day of harvesting, cells were incubated in 22.5 mL of fresh media without serum, and cross-linked by adding 625 μL of 37% formaldehyde (1% final concentration).
Plates were immediately mixed thoroughly after formaldehyde addition, and subsequently rocked every 2 minutes for exactly 10 minutes at RT. The cross-linking was quenched by addition of 1.25 mL of 2.5 M glycine. After 5 minutes at RT, plates were placed on ice for an additional 15 minutes. Cells were finally harvested by a cell lifter, transferred to a 15 mL falcon tubes, and centrifuged at 800 g for 10 min (4ºC). The cell pellet was snap-frozen in liquid nitrogen, and stored at -80ºC until use. Hi-C libraries were then sequenced on a HiSeq4000 (paired end reads of 100 bases each). Two biological replicates of WT and SmcHD1 null MEFs were sequenced in separate lanes, each yielding ~350 million reads per replicate.

Allele-specific alignment
All NGS data were obtained from interspecific FVB-CAST/EiJ cells which enabled allelespecific analysis. Assigning of reads into one of the parental genomes was performed in two stages.
First reads were mapped to mm10 reference genome with the aligner optimal for the assay.
To minimise mapping biases due to differences in the similarity of FVB/Cast genomes to the reference genome all known SNP loci were N-masked before generating appropriate genome index files. SNP coordinates were obtained from The Sanger Institute (ftp://ftpmouse.sanger.ac.uk/REL-1505-SNPs_Indels/). Second using SNPsplit programme 62 and confident SNPs extracted from the file mentioned above reads were sorted into paternal, maternal and unassigned subsets. The procedure in our experimental set up allowed us to assign ~30% reads to parental genomes.

Chromatin RNA-seq
Reads were mapped with STAR 2.5b aligner 63 with index generated from FVB-Cast SNP Nmasked genome. Counts per gene were obtained with Htseq-count 64 .
Initial analysis showed that there is high correlation between chromosome copy number and gene expression values on distinct parental chromosomes therefore HTSeq counts were normalised for chromosomal copy number differences estimated from comparison of ChIPseq input values and repli-seq alignments.
Differential expression analysis was performed with DESeq2 65 , a package from the R Bioconductor project. Results were further processed, analysed and visualized with custom R scripts using commonly used base and Bioconductor packages.

ChIP-seq
ChIP-seq reads were mapped with bowtie2 with SNP N-masked genome index and sorted with SNP-split to separate subsets of reads originating from parental genomes. For SmcHD1, CTCF and Rad21 ChIP-seqs, macs2 was used to detect significant enrichments (narrow peaks, q-value < 0.05). Peak calling was performed on unsorted reads. Allelespecificity of distinct peaks was determined by calculating the ratio of allele-specific reads overlapping the peak in maternal and paternal genome. Peak were considered exclusively specific to either paternal or maternal chromosome if they contained more than 90% of all input corrected reads.
Chromosome-wide SmcHD1 and H3K27me3 ChIP-seq data were obtained by averaging log2(IP/input) values calculated for 500 kb bins within 10 kb intervals.
H3K27me3 depletion domains were defined as regions with intervals of mean log2(IP/input) <0, adjacent regions with negative values separated by no more than 20 kb were merged.

Hi-C
Hi-C reads were mapped with HiCUP pipeline 66  Heatmaps and sums of interaction numbers for specific parts of chromosomes were generated based on balanced Hi-C matrices with HiTC Bioconductor package 69 .

Whole Genome Bisulfite sequencing
Whole Genome Bisulfite libraries were mapped with Bismark 70 , and deduplicated. Further reads were allele-specifically sorted and processed with bismark_methylation_extractor to retrieve methylated cytosines. Cytosines covered by at least 3 reads in allele-specific alignment were used for further analysis.

Repliseq
Repli-seq reads were mapped with bowtie2. Obtained alignments were sorted into parental genome specific subsets and subsequent analysis were performed as described 60 . In brief, reads from S and G1 cell fractions were analysed in 100 kb and 500 kb bins and for each bin read number normalized S/G1 ratio was plotted. Z-scores of S/G1 ratios were further used for detailed analysis and visualization of chromosome-wide replication timing profiles.        The later replication occurs, the lower is the copy number of the DNA sequence. Thus, Sphase DNA copy number profile obtained with high-throughput sequencing (coverage profiles, PCR-free library prep), normalised with an analogous profile for G1 phase from the same cell population, allows the genome-wide replication timing profile to be obtained.