Search
Close this search box.

Differentiation shifts from a reversible to an irreversible heterochromatin state at the DM1 locus – Nature Communications

Deletion of a large CTG repeat expansion from DMPK in DM1 hESCs

To examine whether DMPK hypermethylation could be reversed in DM1-affected human embryonic stem cells (hESCs), we excised a CTG2000 expansion from a mutant hESC line (SZ-DM14, 5/2000 CTGs) with a heavily methylated allele (~100%)14. Using a pair of guide RNAs (gRNAs) designed to target positions −10 bp and +47 bp relative to the repeat sequence (Fig. S1a), we achieved complete removal of the CTG, minimizing alteration in the flanking sequences without altering putative binding sites for CTCF. Validation of repeat deletion from one or both alleles was conducted by Gene Scan analysis (Fig. S1b) followed by a PCR assay overlapping the 5’ breakpoint next to the CTG (Fig. S1c). DNA Sanger sequencing was then utilized to confirm precise repair of the double-strand DNA breaks following repeat excision. Subsequently, selected clones demonstrating successful repeat elimination underwent further validation through Southern blot analysis following enzyme restriction, leaving no room for uncertainty regarding the effective editing of the wild-type and/or mutant allele within the cell (refer to Fig. S2a).

Of the 18 targeted clones established (see Table S1), two were completely DMPK repeat deficient (Δ/Δ). While in one clone the sequence was perfectly repaired with no indels in the normal and expanded alleles (CL9), in the other clone (CL29), one allele was perfectly repaired while the other had a 2 bp deletion at the junction, thus providing compelling evidence that CL9 and CL29 originated from distinctly different clones (Fig. S1d). In the remaining clones, targeting was less efficient and led to a more complex and unpredictable result (Table S1), consistent with our previous findings25. To address concerns about potential off-target effects elsewhere in the genome, we Sanger sequenced candidate sites according to the WTSI Genome Editing tool. The results suggest that no cleavage activity outside of the DMPK locus occurred in both Δ/Δ clones, as corroborated by this assay (Fig. S2b).

Repeat excision reverses aberrant DMPK methylation and heterochromatinization in mutant hESCs

We examined whether the elimination of the expanded repeat had reversed aberrant methylation patterns upstream to the CTGs. To do this, we measured DNA methylation levels before and after repeat excision using bisulfite DNA colony sequencing. Methylation levels were measured nine passages after gene manipulation by colony bisulfite sequencing 650 bp upstream from the CTG repeat (26 CpG sites). This region was previously identified to be part of a disease-associated DMR which is hypermethylated in hESCs when the repeat expands beyond 300 triplets in a way that depends on expansion size14. This experimental approach provided clear evidence for a widespread event of demethylation from 55% (unmanipulated parental cells, corresponding to 100% methylation on the mutant allele) to 0% in both of the Δ/Δ CTG-deficient clones (CL9 and CL29). Additionally, it demonstrated 0% methylation in the unaffected hESC line control (SZ-RB26), both before and after gene editing (Fig. 1a). Locus-specific bisulfite deep-sequencing in an overlapping region (15 CpGs, 759 bp to 631 bp upstream relative to the repeat) in the parental cell line and Δ/Δ clones was used to unequivocally show that methylation was not preserved in any of the successfully edited molecules (Fig. S3a). This contrasted with gene-edited hESC clones CL7 and CL13, in which repeat contraction either resulted in no change, or reduced aberrant methylation levels, respectively (Fig. 1a). This methylation analysis at the DMR in the repeat-targeted clones suggests that DMPK hypermethylation in DM1 depends on the constant presence of the expansion mutation in undifferentiated hESCs.

Fig. 1: Reversal of abnormal methylation and loss of repressive histone modifications by CTG repeat excision in mutant hESCs.
figure 1

a Colony DNA bisulfite sequencing of the DMR (488-777 bp upstream of the repeat, 26 CpG sites) in unmanipulated DM1 hESCs with a heavily methylated CTG2000 expansion (SZ-DM14, methylation levels of 55%), a pair of gene-edited CTG-deficient (Δ/Δ) clones (CL9 and CL29, methylation levels of 0%) and two CRISPRed clones still bearing 1300 CTGs (55%, CL7) and 700 and fewer CTGs (17%, CL13), and wild type hESCs with 5/11 CTG alleles (SZ-RB26, methylation levels of 0%) before and after (Δ/Δ) repeat excision. Filled circles: methylated CpGs; empty circles: unmethylated CpGs. b Real-time PCR ChIP analysis for H3K9me3 in wild type (SZ-13), parental DM1 affected hESC line (SZ-DM14, CTG2000), and isogenic CTG-deficient homozygote clones (CL9 and CL29). APRT and HOXA9 were used as negative and positive controls, respectively. Negative controls were set to one. The data is derived from either n = 3 (SZ-DM14, CL9, and CL29) or n = 4 (wild type) independent ChIP experiments. Each panel illustrates the average ± standard deviation (STD) calculated across all technical replicates. Statistically significant enrichments were calculated within each cell line for DMPK and HOXA9 by comparing to APRT, and between cell lines for DMPK by pairwise comparison to DM1-affected hESC line (two-sided paired t-test). c Real-time PCR ChIP analysis for H3K27me3 in wild type (SZ-13), parental DM1 affected hESC line (SZ-DM14, CTG2000), and isogenic CTG-deficient homozygote clones (CL9 and CL29). APRT and HOXA9 were used as negative and positive controls, respectively. Negative controls were set to one. The data is derived from n = 3 independent ChIP experiments. Each panel illustrates the average ± standard deviation (STD) calculated across all technical replicates. Statistically significant enrichments were calculated within each cell line for DMPK and HOXA9 by comparing to APRT by pairwise comparison to DM1-affected hESC line (two-sided paired t-test). P-values: ns = p > 0.05 *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Precise P-values are provided in Table S4. Source data are provided as a Source data file.

Next, to explore whether hypomethylation is coupled with the loss of heterochromatin, we analyzed the enrichment of two repressive histone modifications: H3K9me3 (representing constitutive heterochromatin) and H3K27me3 (representing facultative heterochromatin) immediately upstream to the DM1 repeat, by chromatin immunoprecipitation (ChIP). While H3K9me3 was exclusively enriched in the unmanipulated DM1 hESCs (Fig. 1b), neither of the hESCs lines/clones were enriched for H3K27me3 (Fig. 1c). Hence, H3K9me3 deposition, but not H3K27me3, is tightly correlated with CTG array size and the gain of aberrant methylation in hESCs. These results indicate that repeat removal in mutant hESCs alters the epigenetic status of the locus in a way that prevents constitutive heterochromatin from being re-established near the repeat.

Altered chromatin structure likely restores CTCF binding but does not affect local gene transcription

It has been suggested that aberrant methylation patterns upstream of the CTGs might alter the chromatin structure through the loss of CTCF binding. Consistent with this claim, we determined methylation levels by bisulfite colony sequencing immediately upstream to the repeat (region F in ref. 14) and validated the binding of CTCF in an overlapping region (CTCF binding site I, CTCFI) by ChIP analysis in wild type, DM1 and Δ/Δ hESCs (Fig. S3b and S3c). Although we were unable to distinguish between the wild type and affected hESCs in terms of extent of enrichment using the ChIP assay (maximum 2-fold change), we found that hypermethylation at the CTCF binding site was exclusive to DM1 hESCs (Fig. S3b). In addition, we leveraged the 2 bp deletion that was induced by gene editing at the junction of one of the alleles in CL29 (unmethylated Δ/Δ clone) to show the presence of two alleles in the CTCF bound fraction after ChIP experiment (Fig. S3d). This, together with the known role of methylation in abolishing CTCF binding next to the repeat in hESCs (Fig. S5D and S5E in ref. 14), and the clear evidence that gene editing does not disrupt CTCF binding next to the CTG (as illustrated by substantial enrichments in Δ/Δ clones, Fig. S3c), strongly suggests that the loss of heterochromatin by repeat deletion restores CTCF occupancy in the mutant allele.

It has been claimed that the change in chromatin structure may alter local gene transcription at the DM1 locus14,26,27. Therefore, we assessed the total mRNA levels of DMPK and SIX5 in the wild type, DM1-affected and DM1-Δ/Δ hESCs by RT-ddPCR. The results showed that the wild type and affected DM1 hESCs did not significantly differ for total DMPK and SIX5 mRNA levels (Fig. S3e). Nor could we find a significant change in SIX5 mRNA levels or a consistent trend in the expression of DMPK, when comparing DM1 unmanipulated vs. gene-edited hESCs (Fig. S3e). Thus, in conjunction with our published data reporting no change in DMPK and SIX5 mRNA expression levels in DM1 unmanipulated as compared to gene-edited myoblasts (see ref. 25), this strongly suggests that local gene expression is unaffected by abnormal methylation at the DM1 locus, at least not in embryonic stem cells and patient myoblasts.

CTG repeat excision does not restore the normal epigenetic status of the locus in patient myoblasts

We explored the relevance of our findings on DMPK demethylation by gene editing to patients’ cells. To do this, we utilized previously established patient-derived repeat-deficient myoblasts25. Since the CTG2600 repeat in these cells was targeted with nearly the same pair of gRNAs as for the hESCs (see Fig. S1a, marked by green asterisks), these cells provided a good opportunity to compare the effect of expanded repeat excision on the methylation status of the DM1 locus in myoblasts vs. undifferentiated hESCs. The analysis of DNA methylation levels in myoblast clones with and without the CTG2600 repeat was carried out precisely as described for the hESCs (i.e. colony bisulfite sequencing 650 bp away from the repeat, 26 CpG sites) after at least 20 population cell doublings. In this case, however, we utilized a non-CpG informative SNP within the DMR to perform allele-specific methylation analysis (rs635299, see also ref. 14). This allowed us to easily distinguish between normal (CTG13; variant G) and expanded (CTG2600; variant T) alleles during the methylation analysis. This approach demonstrated that methylation levels remained unchanged after the complete deletion of the CTG repeat from DMPK in three homozygous (Δ/Δ) and one heterozygous (13/Δ) myoblast clones and presented levels of 100% on the background of the mutant allele (T variant) (Figs. 2a and S4). This was found in addition to the absence of change in methylation levels in three other independent clones, where gene editing was inefficient and failed to remove the CTG repeat from either allele (CTG13/CTG2600). In no case were normal alleles hypermethylated in the wild type control myoblasts or in the mutant myoblasts against the background of the normal allele (variant G, based on the analysis of 150 wild type molecules in total, data not shown), thus ruling out the possibility of non-physiological hypermethylation due to culture conditions.

Fig. 2: CTG excision in affected myoblasts does not restore the normal epigenetic status of the DM1 locus.
figure 2

a Allele-specific colony DNA bisulfite sequencing at the disease-related DMR (488-777 bp upstream of the repeat, 26 CpG sites) pre and post repeat excision (after >20 population cell doublings) from affected myoblasts with normal and expanded alleles (13/2600 CTG), three completely CTG-deficient (Δ/Δ), one heterozygote for the deletion against the mutant allele (13/Δ), three unsuccessfully manipulated clones (13/2600 CTG), and an independent control myoblast cell line (5/14 CTG). Methylation patterns shown for manipulated clones (Δ/Δ, 13/Δ, and 13/2600) on variant T background. Filled circles: methylated CpGs; empty circles: unmethylated CpGs. b ChIP analysis for H3K9me3 in successfully edited (Δ/Δ, clones M4 and M6) versus unsuccessfully edited (13/2600 CTG, clones M1 and M3) DM1 myoblast clones. APRT and MYOGENIN as controls. Data: derived from n = 3 (M4) or n = 4 (M6, M1 and M3) independent ChIP experiments. Each panel illustrates the average ± standard deviation (STD) calculated across all technical replicates. Statistically significant enrichments calculated within each cell line for DMPK and MYOGENIN compared to APRT (two-sided paired t-test). c ChIP analysis for H3K27me3 in successfully edited (Δ/Δ, clones M4 and M6) versus unsuccessfully edited (13/2600 CTG, clones M1 and M3) DM1 myoblasts. APRT as negative control, set to one. Data: derived from n = 2 (M6), n = 3 (M4 and M3) or n = 5 (M1) independent ChIP experiments. Each panel illustrates the average ± standard deviation (STD) calculated across all technical replicates. Statistically significant enrichments calculated within each cell line for DMPK compared to APRT (two-sided paired t-test). P-values: *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Precise P-values are provided in Table S4. Source data are provided as a Source data file.

To explore whether hypermethylation was coordinated with heterochromatin, we confirmed significant enrichment for H3K9me3 by ChIP analysis in all of the affected (successfully and unsuccessfully gene-edited) myoblast clones (Fig. 2b). Strikingly however, when we monitored for H3K27me3, we observed significant enrichment levels in all the examined cell clones, before and after editing. This suggests that H3K27me3 is elicited with heterochromatinization in a way that depends on differentiation into myoblasts. Overall, the results of this analysis lead to the conclusion that the removal of the expansion in DM1-affected myoblasts cannot reset the normal epigenetic status of the DM1 locus once heterochromatin has been established.

Finally, to explore whether persistent methylation in the gene-edited myoblasts could be removed by treatment with a demethylating agent, we monitored for aberrant methylation levels in four different edited clones after a 3 day 5-Aza-dC treatment (5 µM). Strikingly, in three out of the four tested clones, the methylation levels remained unchanged at the DMR (100% against the background of the mutant allele, Fig. S4). In the one remaining clone, methylation levels were reduced from 88% to 55% (M4). None of the methylation levels decreased in the parental DM1 myoblasts as a result of drug treatment (M1). Thus, in most cases (3/4 edited clones), 5-Aza-dC treatment could not restore the normal epigenetic status of the locus against the background of the mutant allele, despite the excision of the repeat.

Differentiation elicits irreversible DMPK hypermethylation in DM1 hESCs

Given the marked differences between the DM1 hESCs and affected myoblasts in terms of the ability to epigenetically reset the locus after gene editing, we next investigated whether differentiation would indeed abolish the reversibility of hypermethylation. For this purpose, we induced the mutant hESCs (SZ-DM14, CTG2000) to spontaneously differentiate in vivo by producing teratomas (benign tumors that contain tissues representatives of all three germ layers) in immuno-compromised mice. These were then used to generate a more homogenous cell population by establishing fibroblast-like cell cultures termed TOFs (teratoma-derived fibroblasts). We chose to target the repeats specifically in TOFs, since the resulting cell cultures are highly homogeneous, can be grown to large numbers, and can easily be transfected, which makes gene editing much more efficient and easier to achieve. Using precisely the same gRNAs as described above for targeting the repeats in hESCs, we removed the CTGs from wild type and expanded alleles (nearly 100% efficiency, see Fig. S5) following cell differentiation. Similar to the results for the edited myoblasts, the excision of the repeats from TOFs did not change the levels of abnormal methylation (45% vs 42%, before and after gene editing, respectively) (Fig. 3a). This provides evidence for the inhibitory effect of differentiation on the reversibility of this process and excludes the possibility of a myogenesis-specific molecular event.

Fig. 3: The shift from a reversible to an irreversible heterochromatin state by hESC differentiation that can be set back by reprogramming after repeat removal.
figure 3

a Colony DNA bisulfite sequencing of the DMR (26 CpG sites) in unmanipulated DM1 affected teratoma-derived cell cultures with a heavily methylated CTG2000 expansion (SZ-DM14 TOFs) before (DM1 Δ/Δ TOFs, 40.9%), and after (DM1 Δ/Δ TOFs CRISPR, 41.6%) repeat excision. Filled circles: methylated CpGs; empty circles: unmethylated CpGs. b A graph summarizing the aberrant methylation levels in iPSCs derived from successfully targeted (M4-IPSC2/3 and M6-IPSC4/5/6, Δ/Δ) and unsuccessfully edited (M1-IPSC1, 13/2600CTG) DM1 affected myoblast clones. Methylation levels at the DMR, against the background of the mutant allele (variant T) were determined by colony DNA bisulfite sequencing (also shown in Fig. S6b), ranging from 0% to 93% in the resulting Δ/Δ iPSCs.

Reprogramming to pluripotency facilitates reversible methylation in affected gene-edited myoblasts

To further substantiate the effect of differentiation on the reversibility of this process, we explored whether the conversion of hypermethylated CTG-deficient myoblasts into iPSCs would restore the normal epigenetic status of the locus. For this purpose, we reprogrammed a pair of myoblast clones that were successfully targeted and accurately repaired (Δ/Δ, clones M4-IPSC2/IPSC3 (derived from M4 edited myoblasts) and M6-IPSC4/IPSC5/IPSC6 (derived from M6 edited myoblasts)) (Fig. S6a). Analysis of DNA methylation levels in the resulting iPSCs was carried out as described for the gene-edited myoblasts by utilizing a non-CpG informative SNP within the DMR (rs635299). This procedure confirmed that reprogramming the gene-edited myoblasts into iPSCs reduced abnormal methylation, albeit to a variable levels in different clones (Fig. 3b and S6b), reaching practically 0% in one of them (M4-iPSC3). This contrasts sharply with the observations in iPSCs with an intact mutation (M1-IPSC1), where methylation levels never change (100% on the background of the mutation). These results demonstrate that the erasure of hypermethylation via the transition to a pluripotent state is conditioned by a preceding step of repeat removal.

Given the extensive cell proliferation of all the investigated cell types (i.e., hESCs, TOFs, myoblasts and iPSCs) it appears very unlikely that de-methylation in the gene-edited pluripotent cells resulted from passive dilution during DNA replication, as opposed to active removal by de-methylating enzymes. Thus, combined with the findings above, we provide evidence for a transition from a reversible to a fixed heterochromatin state at the DM1-expanded locus, which is elicited by differentiation and could be set back by somatic cell reprogramming.

Abnormal methylation at the DM1 locus is maintained by de novo DNMT activity in hESCs

Given the abovementioned results, we aimed to identify chromatin modifiers that would explain the difference in the maintenance of abnormal methylation at the DM1 locus between undifferentiated and differentiated cells.

First, we profiled gene expression by RNA-seq and compared the undifferentiated DM1 hESCs to their differentiated counterparts, teratoma-derived fibroblasts (TOFs). Furthermore, we extended the analysis to include a comparison of undifferentiated hESCs to the patients’ myoblasts, to further substantiate our findings.

After validating the exclusive expression levels of the pluripotent-specific markers POU5F1 (OCT4), NANOG and SOX2 in the hESCs, we compared the expression of nearly 150 potentially relevant genes, collectively representing the complete repertoire of chromatin modifiers in the genome (adapted from dbEM with few additions, see link at http://web.iiitd.edu.in/rghava/dbem/ and ref. 28) (Supplementary Data 1). To visualize the significant differences in the expression of potential chromatin modifiers between the two cell states (hESCs vs. TOFs/myoblasts), Volcano plots were generated (Fig. 4a). Based on this analysis, the de novo DNA methyltransferases (DNMT3a and DNMT3b) and the de-methylating enzymes TET1 and TET2 emerged among the most significant. While DNMT3b and TET1 exhibited exclusive expression levels in the undifferentiated hESCs, DNMT3a and TET2 demonstrated a marked change, with either increased or reduced expression in the hESCs, respectively. This contrasted with the expression patterns of DNMT1 and TET3, both of which maintained comparable mRNA levels across all cell types and states.

Fig. 4: Abnormal methylation at the DM1 locus is maintained by de novo DNMTs activity in hESCs.
figure 4

a Volcano plots comparing chromatin modifier gene expression between undifferentiated DM1 hESCs and their in vivo differentiated counterparts: teratoma-derived fibroblasts (left) or patient myoblasts (right). Red denotes high expression in undifferentiated hESCs, blue in the alternative cell type (TOFs/myoblasts), and gray for equal expression. Green denotes the levels of three pluripotent-associated markers in undifferentiated hESCs. Each data set averages 3 technical experiments. Volcano plot analysis employed edgeR for RNA-seq analysis, with subsequent FDR correction for multiple testing. Plots were generated using the VolcanoNose R program. b DNMT3b targeting approach overview. Bottom: Western blot assesses DNMT3b protein levels in parental DM1-affected hESC line (CTRL) and genetically manipulated isogenic clones, with GAPDH as loading control. The experiment was conducted once. c Residual methylation levels (%) at the DM1-related DMR in single knockouts (SKO) of DNMT3b DM1 hESC clones determined via locus-specific bisulfite DNA deep-sequencing. Levels are relative to baseline in parental hESCs (SZ-DM14), set at 50%. d DNMT3a targeting approach overview. Bottom: DNMT3a mRNA levels assessed before and after gene editing in DNMT3-null DM1-affected hESCs. Residual DNMT3a transcripts post-editing examined by RT-qPCR (exon 17–18, left), and exon 19 skipping validation by RT-PCR (right). CTRL DM1-affected hESCs and M1 myoblasts (myob) served as positive and negative controls, respectively. Data per clone averages n = 4 (I3), n = 5 (3G) or n = 6 (I4, 3K, 3C) technical experiments. Error bars: standard deviation. Significant DNMT3a transcription changes assessed via pairwise comparison (two-sided paired t-test). P-values: ***p < 0.001, ****p < 0.0001. Precise P-values are provided in Table S4. e Residual methylation levels (%) in DM1-related DMR of double-targeted DNMT3a and DNMT3b DM1 hESC clones (DKO) determined via locus-specific bisulfite DNA deep-sequencing. Levels are relative to parental hESCs (SZ-DM14), set at 50%. Source data are provided as a Source data file.

The notable similarity in the differential expression of these enzymes between the two Volcano plots encouraged us to conduct functional assays. First, we chose to knock out DNMT3b, because DNMT3b stood out as the most statistically significant, up-regulated and exclusively expressed chromatin modifier gene of all the hESCs on the list of differentially expressed genes (DEGs) identified (Fig. 4a). By inducing a pair of DSBs with two gRNAs using the CRISPR/Cas9 system, we introduced a homozygous 179 bp deletion overlapping the intron 1-exon 2 boundary in the DNMT3b gene (as depicted in Fig. 4b). We chose to specifically target exon 2, because it is shared by many different mRNA isoforms of the gene (at least 8 different isoforms for DNMT3b). Furthermore, by targeting the 5’-end of the coding region, we increased the probability of introducing premature termination codons (PTCs) or triggering mRNA degradation by nonsense-mediated decay (NMD).

After screening for bi-allelic deletions by PCR (Fig. S7a), the potential knockout clones (KOs) were Sanger sequenced (Fig. S7b) and then validated by Western blot analysis (Fig. 4b). These assays confirmed the complete elimination of the DNMT3b enzyme in 7 out of the 30 transiently selected hESC clones on the background of the DM1 hypermethylated allele.

To evaluate the effect of DNMT3b knockout (KO) on aberrant methylation in these cells, we measured the methylation levels at the disease-associated DMR in three randomly selected clones using bisulfite locus-specific DNA deep-sequencing (at least nine cell passages following gene manipulation). However, the DNMT3b-null clones did not exhibit any discriminable change in aberrant methylation levels when compared to the unmanipulated matched control (Fig. 4c). Given these results, we addressed the possibility of a functional overlap between DNMT3b and DNMT3a enzymatic activity in undifferentiated hESCs, which has been observed in multiple genomic regions29. For this purpose, we targeted DNMT3a on the background of pre-existing DNMT3b-null DM1 hESCs. Using the CRISPR/Cas9 editing system with a pair of gRNAs, we introduced a 128 bp deletion that overlapped with the boundary between exon 19 and the following intron in the DNMT3a gene (Fig. 4d). This approach was chosen for two main reasons. The first is that DNMT3a exhibits multiple isoforms (at least 6) arising from alternative transcription initiation and splicing sites, most of which involve the 3’-end of the gene (exons 7–23 region). The second is that exon 19 is critical for the catalytic activity of the enzyme (exons 16–20), in that it facilitates the comprehensive elimination of potentially active protein species30. Based on these rationales, we introduced a bi-allelic deletion in DNMT3a on the genetic background of DM1 DNMT3b KO hESCs in 5 out of 60 transiently selected puro-resistant clones (Fig. S7c).

After screening for homozygote deletions by PCR and Sanger sequencing (Fig. S7d, e), we performed quantitative reverse transcription PCR (RT-qPCR) to search for a substantial reduction in DNMT3a mRNA levels (Fig. 4d). We found that gene manipulation indeed led to either the complete loss (Cl-I3) or, more commonly, a significant reduction in DNMT3a mRNA levels (Fig. 4d). However, confirming the absence of DNMT3a enzymatic activity in these cells was difficult due to the absence of a straightforward assay for DNMT3a-specific activity. Nonetheless, given the pivotal role of exon 19 in the catalytic domain of DNMT3a30, the residual transcripts were likely to be non-functional because of exon 19 skipping, as validated by Sanger sequencing of the RT-PCR products (Fig. 4d and Fig. S7e). In line with this view, we generated five different bi-allelic DNMT3– double targeted DM1 hESCs.

To assess the effect of de novo DNMT3 double targeting, we monitored the levels of abnormal methylation within the disease-associated DMR by utilizing bisulfite locus-specific DNA deep-sequencing, as described above for single DNMT3b KO clones. There was a significant reduction in abnormal methylation levels in four out of the five assayed clones, ranging from 18% to 38.5% (out of a maximum of 50%, Fig. 4e). Crucially, note that unlike the single DNMT3b KOs, most of the double targeted clones became morphologically abnormal, tended to spontaneously differentiate, and proved to be incapable of sustaining growth beyond five passages.

In summary, this study revealed a fundamental difference between undifferentiated and differentiated cells in terms of the role played by de novo DNMTs (DNMT3a singly or jointly with DNMT3b) in maintaining abnormal methylation patterns at the DM1 locus in undifferentiated hESCs. Furthermore, and unlike in differentiated cells, this molecular event is dependent on the DNA sequence; i.e., the disease-causing CTG expansion at the DMPK gene.