DNA strand asymmetry generated by CpG hemimethylation has opposing effects on CTCF binding

Abstract CpG methylation generally occurs on both DNA strands and is essential for mammalian development and differentiation. Until recently, hemimethylation, in which only one strand is methylated, was considered to be simply a transitory state generated during DNA synthesis. The discovery that a subset of CCCTC-binding factor (CTCF) binding sites is heritably hemimethylated suggests that hemimethylation might have an unknown biological function. Here we show that the binding of CTCF is profoundly altered by which DNA strand is methylated and by the specific CTCF binding motif. CpG methylation on the motif strand can inhibit CTCF binding by up to 7-fold, whereas methylation on the opposite strand can stimulate binding by up to 4-fold. Thus, hemimethylation can alter binding by up to 28-fold in a strand-specific manner. The mechanism for sensing methylation on the opposite strand requires two critical residues, V454 and S364, within CTCF zinc fingers 7 and 4. Similar to methylation, CpG hydroxymethylation on the motif strand can inhibit CTCF binding by up to 4-fold. However, hydroxymethylation on the opposite strand removes the stimulatory effect. Strand-specific methylation states may therefore provide a mechanism to explain the transient and dynamic nature of CTCF-mediated chromatin interactions.


INTRODUCTION
DNA methyla tion pa tterns and their faithful inheritance during cell division are essential for appropriate gene expression ( 1 ).For vertebrates, cytosines in the CpG palindrome in duplex DNA are generally symmetrically methyla ted or unmethyla ted on both strands ( 2 ).An exception to this symmetrical methylation occurs when hemimethylated DNA is generated by DNA synthesis during S-phase.This transitory state of inherent asymmetry is generally ra pidl y converted to symmetrical methylation by DNA methyltr ansfer ase 1 which prefers hemimethylated DNA as a substrate (re vie wed in ( 3 )).Thus, only low le v els of hemimethylation exist in most cells such as somatic mouse tissues ( 4 ).On the other hand, maintained hemimethylation of some repetiti v e sequences has been reported in mouse embryonic stem (ES) cells and in early mouse embryos ( 4 , 5 ).The concept tha t hemimethyla tion is simply a transitory sta te was challenged by the recent discovery of persistent and heritable hemimethylation flanking a subset of CCCTCbinding factor (CTCF) and cohesin binding sites in ES cells, but not in other somatic cell types ( 6 ).The fact that the hemimethylation state is heritable after cell division provides a strong suggestion that it might have biological significance.
CTCF has a major role in genomic organization and function ( 7 ).Also, binding to its recognition motif sequence is well known to be sensiti v e to symmetrical DNA methylation, for example in controlling genomic imprinting of the Igf2 / H19 locus ( 8 , 9 ).We ther efor e used short oligonucleotides (oligos) and recombinant CTCF fulllength (CTCF FL) and truncated proteins containing the ele v en zinc finger (ZF) domains (CTCF ZF1 -11) to investiga te potential dif ferential ef fects of asymmetric DNA methyla tion sta tes on binding in vitr o .Methyla tion on the motif strand inhibited binding; but unexpectedly, methylation on the opposite strand stimulated binding.Potentially, the asymmetry inherent in hemimethylation could be interpreted by CTCF and may serve as a signal for asymmetric cell division in stem cells.

DNA cloning and protein expression
Human CTCF proteins were cloned from the following plasmids: CTCF ZF1-11 (pXC1441) was a gift from Drs Xiaodong Cheng and John Horton ( 10 ) and pDONR223 CTCF WT was a gift from Drs Jesse Boehm, William Hahn, and David Root (Addgene plasmid # 81789; http://n2t.net/addgene:81789; RRID: Ad dgene 81789) ( 11 ).The CTCF ZF1-11 fragment or full-length proteins were cloned into pSUMO vectors for expressing 6xHis-SUMO ta gged CTCF proteins.Ta gged CT CF proteins wer e expressed in the Escherichia coli strain BL21-CodonPlus (DE3)-RIL (230245, Agilent).For large-scale (2 L) purification, a colony was inoculated into 100 mL of LB medium containing ampicillin (50 mg / mL) and 25 M ZnCl 2 and cultured overnight at 37 • C at 170 rpm.Subsequently, the culture was amplified into 2 L of LB medium and was grown for 4-5 h at 28 • C to OD 600 of ∼0.8 and shifted to 16 • C. Protein expression was induced with 0.1 mM isopropyl ␤-D-1thiogalactopyranoside (IPT G) o vernight a t 16 • C .Cells were harvested by centrifugation at 4000 × g for 20 min at 4 • C. Cell pellets were processed immediately or stored at -80 • C for future purification.
For the fluorescence polarization DNA binding assay, both 6xHis-SUMO tagged protein (CTCF ZF1-11) and untagged protein (CTCF ZF1-11 and CTCF full-length) were used, with no change in results based on the presence or absence of the tag.For proteins with the 6xHis-SUMO tag removed, ULP1 Sumo protease was added, and the samples were dialyzed overnight to remove imidazole, then the protein was applied to a 5 ml HisTrap FF column (17525501, Cytiva) pre-equilibrated with buffer A without imidazole.Protein was eluted by a gradient increase in the concentration of imidazole to 250 mM with 10 CVs and subsequently concentrated and loaded onto a Super de x 200 16 / 600 column pre-equilibrated with size-exclusion buffer.The CTCF protein was eluted at a volume of 88 ml for ZF1-11 and 66 ml for full-length protein.The peak fractions were pooled, concentrated, and fresh-frozen in aliquots after dialysis against a size-exclusion buffer containing 50% glycer ol.Pr otein purity and cleavage completion were analyzed by SDS-PAGE.

Site-directed mutagenesis
All site-directed mutagenesis was carried out using the QuikChange kit (200521, Agilent) and the constructs were confirmed by DNA sequencing.

Fluorescence polarization (FP) DNA binding assay
Single-stranded DNA oligos were purchased from IDT and the sequences are shown in Supplementary Table S1.The strand containing the CTCF motif was labeled with 6-carboxy-fluorescein (FAM).Single-stranded oligos were annealed, and the efficiency of the annealing was assessed by running the oligos on a 20% TBE (tris-borate-EDTA) gel.Annealed double-stranded oligos were diluted in DNA binding buffer (20 mM Tris-HCl (pH 7.5), 300 mM NaCl, 5% (v / v) glycerol, and 0.5 mM TCEP) and each oligo (5 nM) was incubated for 15 min at 25 • C with a serial dilution of either full-length CTCF or ZF1-11 CTCF protein.The protein concentration ranged from 4000 to 0.244 nM for full-length CTCF and 8000 to 0.488 nM for ZF1-11 CTCF.Duplicate protein serial dilutions were set up for each oligo in a 384-well black assay plate (#3575, Corning) and the assay was repeated giving a total of 4-8 data sets for each oligo-protein pair.The plates were read on a Synergy Neo micropla te reader (BioTek).Nega ti v e control wells containing only oligo and DNA binding buffer were used as background and subtracted from the polarization values.Polarization ( P ) was then converted to anisotropy ( A ) using the equation ( A ) = (2 × P) / (3 − P ) ( 12 ).Graphpad Prism was used to graph the data and calculate the dissociation constants (K D ) using the nonlinear r egr ession equation for specific binding with Hill slope.A negati v e control oligo was included in each assay with a resulting K D of ≥900 nM.

CTCF ZF3-7 and DNA complex modeling
The CTCF ZF3-7 in complex with DNA structure (PDB ID: 5KKQ ( 10 )) was used as the initial structure template.The DNA fragment was manually mutated to Cen-CTCF or H19 in COOT ( 13 ), and the protein coordinates and DNA coordinates were saved as receptor and ligand, separately.The docking method MDockPP ( 14) was run through the provided w e b server without any constrains.All the other docking parameters were as default.The model was further refined by NPDock DNA-protein complex refinement ( 15 ).

Molecular dynamics simulations
Simulation systems were solvated with TIP3P (tr ansfer able intermolecular potential with 3 points) water with 150 mM NaCl, minimized to convergence by steepest descent.0.1 ns of simulation in the canonical ensemble were then used to relax the solvent and salt with the protein and DNA restr ained, and subsequently equilibr a ted to a tmospheric conditions in the isothermal-isobaric ensemble for 5 ns.All simulations used the GROMACS simulation engine ( 16 ).Production simulations were performed in the canonical ensemble at 300 K.

CTCF binding is regulated by strand-specific CpG methylation
As illustrated in Figure 1 A, there are two key cytosines within the CTCF consensus recognition motif at positions 2 and 12. Wang et al. reported an enrichment of CpG dinucleotides at these two positions (called position 1 and 11 in their paper) and found that 29% of CTCF motifs contain a CpG at one or both of these positions ( 21 ).The C2 and C12 cytosines were shown to interact with the ZF7 and ZF4 domains in the crystal structures reported by Hashimoto et al. ( 10 ).To test the effect of hemimethylation of these cytosines on CT CF binding, four CT CF motifs in the form of 18-19 base pair double-stranded oligos (Figure 1 ) were selected for fluorescence polarization (FP) DNA binding studies ( 12 ).We chose to investigate a CTCF site located at the centromeric boundary of a differentially methylated region (DMR) which includes the only known example of a variably imprinted human gene, nc886 / VTRNA2-1 ( 22 , 23 ).We hav e pre viously shown tha t methyla tion of the nc886 DMR and the resultant repression of nc886 is associated with lower body mass index in children and reduced survival in individuals with acute myeloid leukemia ( 23 , 24 ).Gi v en that the nc886 DMR is variably methylated, we hypothesized that the centromeric CTCF site (Cen-CTCF) would be regulated by methylation.The Cen-CTCF binding motif has CpGs at both the C2 and C12 positions and was discovered using motifbreakR ( https://github.com/Simon-Coetzee/MotifBreakR ( 25 )) to predict whether the rs2346018 A / C polymorphism would affect CTCF binding at this site.The presence of an A provides a consensus nucleotide at position 6 producing a recognition motif for CT CF (Figur e 1 A).The binding of the Cen-CTCF motif was compared to two positi v e control motifs.The first is the CTCF motif located telomeric to the nc886 DMR (Telo-CTCF) which does not contain a CpG site and is frequently occupied by CTCF in multiple cell types based on ENCODE ChIP-seq data.The second positi v e control is a well-studied binding motif from the Igf2 / H19 imprinting control region (H19) which has previously been shown to interact with CTCF ( 8-10 , 26 ).The H19 motif contains a CpG at position C2 and an adjacent CpG in the triplet interacting with ZF6 (Figure 1 A).As a negati v e control, we used the DNA sequence 5 to the Cen-CTCF motif which differs considerably from the consensus sequence and is not predicted to be recognized by CTCF.
For the in vitro binding experiments, we used both fulllength CT CF (CT CF FL) protein and the DNA-binding domain of the protein (CTCF ZF1-11).As shown in Figur e 1 B, ther e was strong binding of both CTCF ZF1-11 and CTCF FL proteins ( K D ∼5 nM) to the positi v e control, Telo-CTCF oligo.The negati v e control showed 200-fold less affinity ( K D ∼1000 nM) establishing the specificity of the assay.The Cen-CTCF and H19 oligos showed intermediate le v els of interaction with both CTCF ZF1-11 and CTCF FL proteins with dissociation constants of about 200 nM.Since the results were similar for both the full-length protein and the DNA-binding portion of the protein and only the ZF domains ar e r equir ed for binding to the motif ( 10 , 26 ), further studies to investigate the effects of DNA methylation on CT CF binding wer e ther efor e conducted with CT CF ZF1-11 protein using the Cen-CTCF and H19 oligos as ligands.
We next measured the effects of various CpG methylation patterns in the Cen-CTCF and H19 oligos on the binding of CTCF ZF1-11 protein using FP assays (Figure 2 ).As expected, full methylation of both strands strongly inhibited binding as shown by an increase in the K D from 163 to 1075 nM (7-fold differ ence) (Figur e 2 A).Methylation of only the motif strand C2 and C12 positions had an almost equivalent effect as full methylation of both strands ( K D = 1034 nM).On the other hand, methylation of the opposite strand C2 and C12 positions increased binding to a K D of 37 nM showing a 28-fold difference in binding between the two hemimethyla tion sta tes (Figure 2 A).Further investiga tion of the roles of individual C2 and C12 methylation and combinations of methylation showed the dominant role of motif strand C2 methylation in inhibiting binding and the additi v e nature of opposite strand methylation in stimulating binding (Supplementary Tab le S2).Gi v en the surprising nature of these observations, we verified the stimulatory effects of hemimethylation using the H19 oligo, since full methylation of this site has been well studied ( 8 , 9 ).As indicated in Figure 1 A, this motif has a CpG at the C2 but not the C12 position.Full methylation of the C2 position on both strands or methylation of only the motif strand C2 position changed the K D from 149 nM to 294 and 346 nM respecti v ely (Figure 2 B).Once again, methylation of the opposite strand stimulated binding ( K D 89 nM), and the difference between the two hemimethylation states was 4-fold (Figure 2 B).
Two previous studies assessed hemimethylation of CTCF binding sites within the imprinting control region upstream of the mouse and human H19 gene but did not report a change in CTCF binding when the opposite DNA strand was methylated ( 9 , 26 ).One of these studies included the MS1 site from the mouse HS1 insulator ( 9 ) which contains the same motif sequence used in the H19 oligo in this paper.The gel mobility shift assay that was used in these previous studies may not have been sensitive enough to observe or quantitate an increase in binding.Here, we find that the effects of strand-specific methylation depend on the actual CTCF target sequence; and the inhibition of binding from motif strand methylation is dominant over the stimulation of binding from opposite strand methylation.

Hemimethylated CpGs occur within CTCF motifs in embryonic stem cells
Our results that CTCF binding can be regulated by strandspecific hemimethyla tion sta tes led us to investigate whether any of the hemimethylated CpGs identified in H9 human embryonic stem cells by Xu and Corces (2018) ( 6 ) and in undif ferentia ted E14TG2a mouse embryonic stem cells by Zhao et al. ( 17 ) are located within a CTCF motif.In the Xu and Corces paper (2018), the authors found that hemimethylated CpGs often flank CTCF motifs and show an asymmetry as to which strand is methylated ( 6 ).We used their nascent bisulfite sequencing pulse-chase data to determine whether CpGs within the recognition motif itself are hemimethylated.Using the CTCF consensus sequence data sources as described in Dozmorov et al. ( 18 ), we identified 102838 CpGs that fall within a CTCF motif of which 803 of these CpGs were found to be hemimethylated across both pulse and chase using a 50% difference in methylation cutoff (Table 1 and Supplementary Table S3).Using the hairpin bisulfite sequencing data from the Zhao et al. (2014) paper, we found 189291 CpGs that fall within a CTCF motif of which 2604 are hemimethylated using a 50% difference in methyla tion cut-of f (Table 1 and Supplementary Table S4).It is worth noting that the Cen-CTCF motif is not present in the mouse genome.In the human ES cell line, the Cen-CTCF motif does appear to contain hemimethylation; howe v er, it was not included in our final data set because we limited our results to CpGs with hemimethylation across both pulse and chase and there was missing data at this site.The data from both studies provide evidence that about 1% of CpGs within CT CF motifs ar e hemimethylated in vivo , and there is a pproximatel y an e v en distribution of methylation on either the motif strand or the opposite strand in both mouse and human ES cells (Table 1 ).This suggests tha t strand-specific hemimethyla tion could potentially have opposing effects on CTCF binding in vivo as predicted by our in vitro data, but further detailed studies in living cells are needed to substantiate this.
To assess the possible biological significance of these hemimethylated CpGs, we used gene ontology (GO) enrichment analysis of the genes which contain or are near the hemimethylated CpGs.We found that hemimethylation at CTCF motifs is enriched within / near genes involved in tissue de v elopment and cellular dif ferentia tion.The presence of hemimethylated CpGs on the opposite strands may str engthen CT CF binding and provide a mechanism to maintain CTCF at certain critical sites during cell division.
Ther efor e, it is possible that this mechanism could be involved in regulating the dif ferentia tion program in stem cells.

CTCF V454 and S364 sense the presence of methyl groups on the opposite strand
To explore the structural basis for the surprising increase in binding of CTCF to methylation on the opposite strand, we utilized the detailed crystallo gra phic data of Hashimoto et al. (PDB ID: 5T00 ( 10 )), which was obtained using a differ ent CT CF motif containing a CA instead of a CpG at the C2 position with the resultant thymine rather than a cytosine on the opposite strand.This opposite strand thymine presents a methyl group in exactly the same geometry as 5-methylcytosine (5mC) and was shown to have hydrophobic interaction with V454 of ZF7 in the existing crystallo gra phic data (PDB ID: 5T00 ( 10)).In addition, the C beta from the S364 of ZF4 has a hydrophobic interaction with 5mC at the C12 position on the opposite strand.In agreement with this observation, our Molecular Dynamics (MD) simulations using the 5T00 structure and substituting in the Cen-CTCF oligo re v ealed that V454 and S364 remain in close proximity to the mCpG at positions C2 and C12 on the opposite strand for hundreds of nanoseconds (Supplementary Figure S1).This observation led us to hypothesize that V454 of ZF7 and S364 of ZF4 sense opposite strand methylation through hydrophobic contact with 5mC.Docking of the opposite strands of the C2 and C12 positions of Cen-CTCF was performed in silico to investigate the potential interactions between these key cytosines and ZF7 and ZF4 (Figure 3 A and B).The Valine residue at position 454 and the Serine residue at position 364 in the protein were found to be in close proximities to the C2 and C12 opposite nucleotides respecti v ely, and ther efor e may interact with the methyl group of the 5mC through hydrophobic interactions.Molecular modeling predicted that the addition of a methyl group to the C2 opposite position would decrease the interaction distance from 4.9 angstroms ( Å ) to 3.5 Å and from 5.1 to 3.6 Å for the C12 opposite position (Figure 3 A and B).Similar r esults wer e found for the C2 position on the H19 motif where 5mC decreased the interaction distance from 4.2 to 3.0 Å suggesting that this mechanism might be general (Supplementary Figure S2).The likelihood that these predicted interactions with 5mC would increase the binding of CTCF was investigated by exploring the effects of substituting the key residues (V454 and S364) in the protein with glycine (hydrophobic interaction reducing) and investigating the influence of 5mC on binding to Cen-CTCF.As expected, methylation of the motif strand inhibited binding by all mutant proteins (Figure 3 C-F).On the other hand, the stimulatory effect of 5mC on the opposite strand was reduced by substitution to glycine at either the V454 or S364 positions (Figure 3 C and D), but not at the S450 or V363 positions (Figure 3 E and F).These data strongly support the structural predictions by show-ing that V454 and S364 residues in the CTCF ZFs sense the presence of methyl groups on the opposite strand C2 and C12 positions.

CTCF binding is regulated by CpG h y droxymeth ylation
Se v eral studies hav e reported changes in 5h ydroxymeth ylcytosine (5hmC) le v els at CTCF binding sites (27)(28)(29)(30); howe v er, little is known about the effects of 5hmC on CTCF binding.The 5hmC base is predicted to not only introduce a hydrophilic group, but also to sterically obstruct the interactions between the CTCF V454 and S364 with the opposite strand C2 and C12 positions of Cen-CTCF and C2 of H19 (Figure 4 A and C).Not shown, is the prediction that 5hmC on the motif strand would still inhibit binding gi v en that it is larger than 5mC.We tested these predictions using the Cen-CTCF and H19 oligos containing 5hmC at the target positions (Figure 4 B and D).Full h ydroxymeth ylation of the CpGs within the Cen-CTCF oligo or h ydroxymeth ylation on the motif strand resulted in a 4-fold reduction in the binding affinity (Figure 4 B), confirming that 5mC or 5hmC at the C2 position sterically obstructs D451 in CTCF ( 10 ).The presence of 5hmC on the opposite strand had no effect or subtle effects on the binding of CT CF (Figur e 4 B).For the H19 oligo, h ydroxymeth ylation on both strands of the oligo or only on the motif strand resulted in a 2-fold reduction in the binding affinity (Figure 4 D).As with the Cen-CTCF oligo, h ydroxymeth ylation of the opposite strand of the H19 oligo had only a subtle effect with slightly diminished binding af finity rela ti v e to the unmethylated oligo (Figure 4 D).Ther efor e, the substitution of 5hmC for 5mC on the opposite strand mitigated the stim ulation, w hereas it maintained the inhibitory effects on the motif strand, albeit to a lesser extent than 5mC.Our finding that hydroxymethylation can inhibit the binding of CTCF to DNA is in line with a study by Marina et al. (2016) in which they show by electrophoretic mobility shift assay that DNA probes from the CD45 and KCNA2B genes with the modifications 5mC , 5hmC , and 5fC have minimal interaction with CTCF ( 31 ).Variable h ydroxymeth ylation might ther efor e add an additional le v el of control to CT CF binding.Pr evious reports have shown that h ydroxymeth ylation is increased at weak and variable CTCF sites ( 29 ) and have suggested a role for h ydroxymeth ylation in maintaining a poised chroma tin sta te ( 28 , 29 ).Thus, h ydroxymeth ylation may r epr esent a dynamic intermediate that maintains the motif in a poised state ready for changes in cytosine modifications that could stimulate the binding of CTCF.

CONCLUSIONS AND IMPORTANCE
The recent discovery of heritable hemimethylation states in ES cells ( 6 ) has challenged the long-held assumption that hemimethylation is simply an intermediate to full symmetrical methylation and ther efor e of little biological significance.Our results tha t dif ferent hemimethyla tion sta tes a t CTCF binding sites can change binding affinity by as much as 28-fold suggests that there may indeed be a function for hemimethylation in the organization of the genome, particularly in ES cells.We were helped considerably in defining the mechanism for these effects by the detailed structures published by Hashimoto et al. ( 10 ) and were able to confirm the mechanism by mutating the relevant zinc fingers in CTCF.Our results are particularly intriguing because hemimethyla tion a t a single CTCF motif could potentially cause large-scale alterations of the epigenome in ES cells.A recent study by Gabriele et al. demonstra ted tha t e v en CTCF topolo gicall y associated domain boundaries, which are thought to be stable interactions, can be highly dynamic and transient in natur e ( 32 ).The r egulation of CT CF sites by specific CpG modified states in a strand-specific manner, like hemimethylation and hydroxymethylation, may provide a mechanism to explain the transient and d ynamic na ture of CTCF-media ted chroma tin interactions.
The phenomenon of asymmetric cell division was identified by Edwin Conklin > 100 years ago ( 33 ) and many mechanisms likely participate in the generation of cellular di v ersity ( 34 ).While there is no evidence yet that DNA methylation plays a role in this process, hemimethylation may provide a mechanism for asymmetric cell division given that it generates two sequence-identical DNA duplexes with strongly differing potentials to bind CTCF.In this regard, it is interesting that hemimethylation is much less common in the dif ferentia ted progeny of ES cells ( 6 ) suggesting that resolution of hemimethylation may be associated with cell dif ferentia tion.

DA T A A V AILABILITY
All data are available from the corresponding author upon request.

SUPPLEMENT ARY DA T A
Supplementary Data are available at NAR Online.

Figure 1 .
Figure 1.CTCF binds to the Cen-CTCF motif.( A ) Alignment of the oligo motif strand DNA sequences to the CTCF consensus sequence from motifbreakR and to CTCF zinc fingers (ZF) 3-7 as reported by Hashimoto et al. ( 10 ).Bold = consensus nucleotide, red = mismatch from the consensus sequence, underline = CpG.( B ) Binding affinity curves and dissociation constants ( K D ) of CTCF ZF1-11 and CTCF FL proteins for the oligos defined in (A).Double-stranded oligo sequences are shown with the core motif colored orange for Telo-CTCF, blue for Cen-CTCF, and green for H19.Binding data are r epr esented as mean ± SEM, n = 4. # : The calculated K D of Telo-CTCF is at or near the probe concentration of 5 nM, indicating very tight binding ( K D ≤ 5-6 nM).

Figure 2 .
Figure 2. CTCF binding is regulated by strand-specific CpG methylation.( A to B ) Binding affinity curves and dissociation constants ( K D ) of CTCF ZF1-11 for the Cen-CTCF (A) and H19 (B) oligos methylated at the indicated positions.Binding data are represented as mean ± SEM, n = 4 (Cen-CTCF) and n = 8 (H19).Fold change in K D between different methylation states is indicated using a bracket.The core motif sequence within the oligo is colored blue.

Figure 3 .
Figure 3. CTCF V454 and S364 sense the presence of methyl groups on the opposite strand.Close-up of V454 ( A ) and S364 ( B ) interactions with opposite strand C2 and C12 positions in the molecular modeling of CTCF ZF3-7 with the unmodified Cen-CTCF motif (left), and opposite stand CpG methylated Cen-CTCF motif (right).Modeling a methyl group onto C2 and C12 opposite positions increases the hydrophobic interaction with V454 of ZF7 and S364 of ZF4.Proteins are in cartoon presentation.Specific nucleotides and key valine and serine residues are in stick presentation.Atoms are colored grey for carbon, blue for nitr ogen, br own for phosphorus, and red for oxygen.The 5-methyl groups are decorated with spheres.The numerical numbers indicate the inter-atomic distance in angstroms.(C-F) Binding affinity curves and dissociation constants ( K D ) of CTCF ZF1-11 valine-to-glycine and serine-to-glycine mutant proteins for the Cen-CTCF motif methylated at the indicated positions.Mutant residues V454G ( C ) and S450G ( E ) are in close proximity to the C2 position on the opposite strand, and mutant residues S364G ( D ) and V363G ( F ) are in close proximity to the C12 position on the opposite strand.Binding data are represented as mean ± SEM, n = 4. Fold change in K D between different methylation states is indicated using a bracket.

Figure 4 .
Figure 4. CTCF binding is regulated b y CpG hy droxymethylation.( A ) Modeling of a hydr oxymethyl gr oup on the opposite strand C2 and C12 positions of the Cen-CTCF oligo w hich potentiall y results in repulsion (indicated by stars) with V454 of ZF7 and S364 of ZF4 in CTCF.( B ) Binding affinity curves and dissociation constants ( K D ) of CTCF ZF1-11 for the Cen-CTCF oligo h ydroxymeth yla ted a t the indica ted positions.( C ) Modeling of a h ydroxymeth yl group on the opposite strand C2 position of the H19 oligo potentially results in repulsion (indicated by stars) with V454 of ZF7 in CTCF.( D ) Binding affinity curves and K D s of CTCF ZF1-11 for the H19 oligo h ydroxymeth yla ted a t the indica ted positions.Binding da ta are represented as mean ± SEM, n = 4 (Cen-CTCF motif) and n = 8 (H19 motif).Fold change in K D between different methylation states is indicated using a bracket.The core motif sequence within the oligo is colored blue.For the modeling images, proteins are in cartoon presentation.Specific nucleotides and valines are in stick pr esentation.Atoms ar e color ed gr ey for carbon, blue for nitr ogen, br own f or phosphorus, red f or o xygen.The hydro xymethyl groups are decorated with r ed spher es.

Table 1 .
Hemimethyla ted CpG d yads within CTCF motifs exist in embryonic stem cells Human nascent bisulfite sequencing pulse-chase hemimethylation data from Xu and Corces (2018) and mouse hairpin bisulfite sequencing hemimethylation data from Zhao et al. (2014) were downloaded from GEO database (GSE97394 and GSE48229).* Only CpG dyads with hemimethylation across both pulse and chase were included.