Dppa2/4 Counteract De Novo Methylation to Establish a Permissive Epigenome for Development

Early mammalian development entails genome-wide epigenome remodeling, including DNA methylation erasure and reacquisition, which facilitates developmental competence. To uncover the mechanisms that orchestrate DNA methylation (DNAme) dynamics, we coupled a single-cell ratiometric DNAme reporter with unbiased CRISPR screening in ESC. We identify key genes and regulatory pathways that drive global DNA hypomethylation, and characterise roles for Cop1 and Dusp6. We also identify Dppa2 and Dppa4 as essential safeguards of focal epigenetic states. In their absence, developmental genes and evolutionary-young LINE1 elements, which DPPA2 specifically binds, lose H3K4me3 and gain ectopic de novo DNA methylation in pluripotent cells. Consequently, lineage-associated genes (and LINE1) acquire a repressive epigenetic memory, which renders them incompetent for activation during future lineage-specification. Dppa2/4 thereby sculpt the pluripotent epigenome by facilitating H3K4me3 and bivalency to counteract de novo methylation; a function co-opted by evolutionary young LINE1 to evade epigenetic decommissioning.

ratiometric DNAme reporter with unbiased CRISPR screening in ESC. We identify key genes and 6 regulatory pathways that drive global DNA hypomethylation, and characterise roles for Cop1 and 7 Dusp6. We also identify Dppa2 and Dppa4 as essential safeguards of focal epigenetic states. In their 8 absence, developmental genes and evolutionary-young LINE1 elements, which DPPA2 specifically 9 binds, lose H3K4me3 and gain ectopic de novo DNA methylation in pluripotent cells. Consequently, 10 lineage-associated genes (and LINE1) acquire a repressive epigenetic memory, which renders them 11 incompetent for activation during future lineage-specification. Dppa2/4 thereby sculpt the pluripotent 12 epigenome by facilitating H3K4me3 and bivalency to counteract de novo methylation; a function co-13 opted by evolutionary young LINE1 to evade epigenetic decommissioning. 14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  INTRODUCTION   49  50 Mammalian fertilisation is accompanied by widespread epigenetic remodeling of inherited genomes, 51 including global DNA demethylation and reorganization of chromatin landscapes 1-4 . This epigenetic 52 resetting equalises the distinct parental epigenomes, and also correlates with the emergence of naïve 53 pluripotency, implying epigenome remodeling is central to establish developmental competence. Such 54 'competence' confers the capacity of the genome to transcriptionally respond to future inductive signals 55 for multiple lineages, and can be considered the execution of pluripotency. This is particularly critical 56 for lineage-associated genes that need to be transiently repressed during pluripotent phases, whilst 57 remaining competent (primed) for robust activation in subsets of forthcoming cell fates 5 . Indeed, the 58 importance of a permissive epigenome is supported by observations of impaired or reduced 59 developmental competence after somatic cell nuclear transfer (SCNT) or in induced pluripotent stem 60 (iPS) cells, which are susceptible to incomplete epigenetic resetting 6,7 . Investigating the complex 61 mechanisms that underpin epigenome (re)programming is therefore an important focus towards 62 understanding developmental potency. 63 64 Several lines of evidence indicate that resetting DNA methylation (DNAme) during development is 65 mediated by parallel mechanisms 8 . Amongst these, repression of the maintenance DNA methylation 66 machinery is central and appears to occur through post-translational regulation of UHRF1 9,10 , at least 67 in part via STELLA activity 11,12 . This is further supported by PRDM14, which suppresses the de novo 68 methylases, and is necessary for DNA hypomethylation in naïve pluripotent cells 13,14 . In parallel, 69 replication-independent DNAme erasure occurs on both the maternal and paternal genomes 1 . 70 Counterintuitively, de novo methylation remains active throughout epigenetic reprogramming but is 71 offset, in part, via TET proteins 15 . These collective mechanisms contribute towards resetting the 72 epigenome, but also present an opportunity for transposable elements (TE), such as LINE1, to mobilise 73 due to epigenetic de-restriction. Such LINE1 activation has been linked with key developmental 74 events 16 , but could also represent a hazard to the genome if left unrestrained 17,18 . Epigenetic 75 (re)programming therefore likely strikes a balance between genome-wide resetting to a competent state 76 for development, and targeted regulation. Nevertheless, a complete understanding of the mechanisms 77 that cross-talk to remodel the epigenome, how they interact to balance focal and global effects, and 78 what the full repertoire of genes involved is lacking. 79 80 Here we have coupled a single-cell ratiometric reporter of cellular DNA methylation status with 81 CRISPR screening to unbiasedly identify the gene networks that underpin DNAme remodeling. In 82 doing so we identify upstream regulators of global DNAme erasure in pluripotent cells. We also identify 83 Dppa2 and Dppa4 as key genes that safeguard against focal de novo DNA methylation and epigenetic 84 silencing at lineage-associated genes by integrating chromatin states, and consequently confer 85 developmental competence. Remarkably, LINE1 elements appear to have exapted this Dppa2/4 86 function to escape epigenetic surveillance and enable competence for precocious activation, potentially 87 highlighting an evolving genomic conflict.

92
Single-cell monitoring of DNA demethylation 93 To identify regulators of epigenetic remodeling we exploited the reporter for genomic DNA methylation 94 (RGM) 19 , which tracks the dynamic DNA methylation state of single-cells with GFP. We optimised the 95 system for CRISPR screening in two ways. First, we replaced the original Snrpn imprinted promoter 96 for the core Kcnq1ot1 imprinted promoter, which enhanced the dynamic range of reporter activity 97 (eRGM), enabling better separation of hypo-and hyper-methylated cells ( Fig S1A). Second, we 98 converted the read-out to a ratiometric measure by introducing an additional Ef1a-mCherry that is not 99 sensitive to DNA methylation ( Fig 1A). This enables a single-cell ratiometric score (eRGM(GFP):Ef1a-100 mCherry) that normalises for general confounding effects on a reporter in a screen (e.g. disruption of 101 translation factors) or inherent cell-cell variance (e.g. cell cycle stage).

103
To test ratiometric eRGM we developed a model of developmentally induced DNA demethylation. 104 Here, murine embryonic stem cells (ESC) are maintained in a titrated 2i/L (t2i/L) condition (see 105 methods) to promote high global levels of DNA methylation (range: 64%-58%), and are then 106 transitioned to 2i/L status to induce global demethylation (range: 30%-44%; p=0.0002) (Fig 1b). 107 Importantly, global DNA demethylation after switching from t2i/L→2i/L occurs without significant 108 changes in cell identity, as judged by transcriptome, which is in contrast to the switch from conventional 109 serum/LIF to 2i/L that constitutes a major transcriptional shift 20-22 (Fig 1B & S1B). Moreover, the 110 induced DNA hypomethylation pattern is well correlated with developmentally imposed DNA 111 demethylation in vivo ( Fig S1C). Thus, the t2i/L→2i/L model specifically captures an authentic global 112 epigenetic transition, including global DNA demethylation, without changes in cell identity that could 113 confound a screen for epigenome regulators. 114 115 We next examined the capacity to detect DNA demethylation events in single-cells by generating 116 independent ESC lines carrying the ratiometric eRGM system. In t2i/L eRGM was silenced in >95% 117 of cells, consistent with high global DNA methylation. In contrast, eRGM exhibited a progressive 118 activation concomitant with induced DNA methylation erasure in 2i/L, leading to eRGM activation in 119 12% of single-cells after 3 days, 67% after 6 days, and in >95% of cells upon complete DNA 120 hypomethylation at 12 days (Fig 1C & S1D). Independent eRGM lines exhibited consistent response to 121 induced hypomethylation ( Fig S1D). Notably, Ef1-mCherry did not alter expression during this 122 transition enabling its use as a ratiometric normaliser (Fig S1D). To further confirm eRGM directly 123 reports cellular DNA methylation status, we used ESC wherein tamoxifen (TAM) drives CRE-mediated 124 deletion of Dnmt1 (cDKO) and, consequently, global DNA demethylation occurs independent of 125 culture condition 23 . Upon TAM exposure we observed a strong and progressive activation of eRGM 126 amongst single-cells concomitant with Dnmt1 (cDKO)-induced DNA hypomethylation ( Fig 1D).
127 128 Finally, we tested whether eRGM can also respond reciprocally to acquisition of DNA methylation by 129 inducing differentiation of hypomethylated ESC (in 2i/L) into hypermethylated EpiLC (global 5mC 130 33%→75%). Here the reporter initiated rapid silencing in parallel with induction of DNA 131 hypermethylation ( Fig 1E). We conclude the enhanced ratiometric reporter of genomic DNAme 132 (eRGM) represents a single-cell read out for dynamic transitions of cellular DNA methylation status. 133 134 135 A CRISPR screen for regulators of dynamic DNA methylation 136 To identify critical factors for DNA methylation resetting, we generated independent ESC lines carrying 137 ratiometric eRGM and a single-copy of spCas9, and introduced into them a CRISPR knockout gRNA 138 library 24 . To validate the capacity of the strategy to detect epigenetic regulators, we isolated ESC that 139 activated eRGM under hypermethylated (t2i/L) conditions, which is predicted to identify factors 140 necessary to maintain DNA methylation and/or epigenetic silencing. Analysis using MAGeCK 25 , 141 revealed the top hits comprised the key machinery for maintenance DNA methylation, in particular 142 Dnmt1 (rank: 5, FDR=0.00049) and Uhrf1 (rank: 48, FDR=0.066), unbiasedly confirming eRGM 143 sensitivity to DNA hypomethylation ( Fig S1E). We also identified regulators of chromatin-mediated 144 silencing such as Setdb1 (rank: 51, FDR=0.073,) and the HUSH complex (Mphosph8 rank: 6; Morc2a 145 rank: 9; Fam208a; rank: 13). These data support eRGM specificity for detecting developmental 146 epigenome regulators, including of cellular DNA methylation status.

148
We next aimed to identify factors that contribute to resetting the epigenome at focal or global scales. 149 We induced global DNA demethylation and isolated individual ESC that failed to ratiometrically 150 activate eRGM, indicative of a failure to undergo epigenetic resetting ( Fig 1F). Importantly, this 151 population was highly enriched for knockout of Prdm14 (rank: 16, FDR=0.0006) the key regulator 152 known to instruct global DNA demethylation 13 as well as its heterodimeric co-factor Cbfa2t2a (rank: 153 20, FDR=0.0006) 26,27 , supporting the sensitivity of the strategy for identifying reprogramming factors 154 ( Fig 1F). Moreover, screens of independent eRGM ESC lines identified highly correlated (p=0.01, 155 spearman RRA) candidates ( Fig S2A) suggesting the system is robust. We therefore intersected 156 significant hits (FDR<0.05, FC>3) from independent screens to identify a core candidate list of 56 157 putative genes linked with resetting the epigenome (Table S1). To validate the CRISPR screen hits we generated knock-out (KO) ESC populations for 24 selected 173 candidates, by introducing a constitutive gRNA into eRGM:Cas9 lines via piggyBac, and transited to 174 hypomethylated conditions. Strikingly, knockout of each candidate resulted in a degree of impaired 175 eRGM activation, implying altered epigenome remodeling in their absence (Fig 2A). This effect was 176 robust since we generated additional knockouts in an independent eRGM line, with similar outcomes 177 ( Fig S2C). Interestingly, the response kinetics of eRGM during transition to 2i/L varied amongst 178 candidate KO. For example, Jak1, Dppa2, Dppa4 and Brd4 mutants failed to activate eRGM per se, 179 indicating a general block. In contrast, other candidate knockouts such as Dusp6, Kdm3a,Nufip1 and 180 Cop1 exhibited late-onset heterogeneous activation amongst single-cells, implying delayed 181 demethylation dynamics and reduced robustness in their absence. (Fig 2B & S2D). These validations 182 suggest that candidate factors influence both the kinetics and absolute response of eRGM.

184
To test whether the impaired eRGM response in candidate KO is directly indicative of incomplete 185 epigenetic reprogramming, we used LUMA to quantitatively assess global DNA methylation. 186 Consistent with eRGM, we found knockout of 20 of 24 candidate factors resulted in impaired global 187 DNA demethylation across independent lines ( Fig 2C). Amongst these is the known epigenetic 188 regulator Prdm14, which maintained 64-58% global DNAme relative to hypomethylated WT control 189 (39%), as well Cbfa2t2 (54-52%). Novel candidates that exhibited substantially elevated DNAme upon 190 knockout and transition to 2i/L include the phosphatase Dusp6 (60-56%), the tyrosine kinase Jak1  65%), the epigenetic regulator Brd4 (59-51%), and the E3 ubiquitin ligase Cop1 (56-54%). These data 192 suggest our screen is sufficient to identify critical components of gene regulatory networks that 193 contribute to driving complete DNA demethylation in naïve ESC. 194 195 196 Dusp6 and Cop1 promote global DNA hypomethylation 197 To further investigate the role of candidates Dusp6 and Cop1 in epigenetic transitions, we generated 198 independent clonal knockout ESC lines. DUSP6 is a phosphatase that acts downstream of MEK to 199 attenuate the ERK signal cascade, whilst COP1 mediates ubiquitination and proteasomal degradation 200 of target proteins 29,30 . We used enzymatic methyl(EM)-seq, an enhanced equivalent of bisulfite (BS)-201 seq, to chart the global DNA methylome in WT, Dusp6 -/and Cop1 -/naïve ESC, which confirmed 202 mutant lines remain hypermethylated in in 2i/L (Dusp6 -/-67%; Cop1 -/-58%) ( Fig 2D). Notably, elevated 203 DNAme is distributed equivalently across genomic features such as promoters, repeats and intergenic 204 regions, indicating a general impairment to DNA demethylation rather than failure in locus-specific 205 resetting (Fig 2D-E suggesting a general acquisition of focal hypermethylation in Dppa2 -/-ESC ( Fig 3C). Dppa4 -/-ESC 240 exhibit a highly correlated pattern of hypermethylated DMRs as Dppa2 -/- (Fig 3C), likely because 241 disruption of one protein led to reciprocal destabilization of the other (Fig S3A). Notably, DNA 242 hypermethylation was apparent at many gene promoters that usually remain strictly unmethylated at all 243 developmental stages ( Fig 3D). This indicates that rather than impaired DNA demethylation per se in 244 Dppa2/4 mutants, there is aberrant de novo methylation activity that could establish DNA methylation 245 'epimutations'. 246 247 To investigate this further, and determine whether such epimutations persist during differentiation, we 248 profiled epiblast-like cells (EpiLC), which correspond to a formative state that has undergone epigenetic 249 programming (genomic re-methylation). We observed the hypermethylated sites established in Dppa2 - Because the DMRs are focal rather than global, we next asked whether they reflect localised DPPA2/4 267 activity. To test this, we performed CUT&RUN for DPPA2 binding in WT cells and observed genomic 268 occupancy is strikingly increased over sites that become hypermethylated DMR in Dppa2/4 -/cells, 269 suggesting DPPA2 may act proximally to sculpt the DNA methylome ( Fig 3F). Indeed, DPPA2 binding 270 peaks (n=28,338) are significantly enriched specifically over gene promoters (p<0.001) and the 5' end 271 of full-length LINE1 elements (>5kb) (p<0.001) (Fig 3G), consistent with DMR associations. Overall, 272 promoters and LINE account for >65% of DPPA2 genomic occupancy. The latter enrichment is specific 273 for full-length LINE since DPPA2 is not enriched at truncated LINEs or other repetitive genomic 274 features ( Fig 3H). Amongst promoters, DPPA2 exhibits a preference for CpG-dense loci, which is 275 reflected by its GC-rich binding motifs and preference for CpG island (CGI) promoters (Fig 3I & S3E). 276 Notably, DPPA2 binding was observed at the sensor region used for eRGM (Fig S3F), explaining why  To understand the broader chromatin features associated with DPPA2 occupancy, and susceptibility to 285 hypermethylation, we used CUT&RUN to profile H3K4me3, H3K27me3, and H3K9me9. Notably, we 286 observed that DPPA2 occupancy directly correlates with strong H3K4me3 enrichment in WT cells, 287 across all binding sites ( Fig S4A). H3K27me3 is also enriched at a subset of DPPA2-bound sites, 288 establishing bivalent states, but H3K9me3 is largely depleted. Strikingly, the H3K4me3 enrichment at 289 DPPA2-bound promoters occurs irrespective of expression state, in both ESC and EpiLC (Fig 4A), 290 implying that DPPA2 may directly target H3K4me3, rather than H3K4me3 reflecting expression status 291 of DPPA2-bound sites. Importantly, loci that acquire hypermethylation in Dppa2/4 mutants (DMR) 292 correspond to genomic regions that are H3K4me3-enriched and DPPA2-bound in WT ( Fig S4B). Taken 293 together this suggests a potential connection between DPPA2 occupancy, H3K4me3 and DNA 294 methylation status. 295 296 To investigate this further we assayed H3K4me3, H3K27me3 and H3K9me3 in Dppa2 -/and Dppa4 -/-297 ESC and EpiLC. Remarkably, deletion of Dppa2 or Dppa4 results in a dramatic loss of H3K4me3 298 across a significant subset of DPPA2-bound sites in both ESC and EpiLC, whilst the remaining sites 299 are apparently unaffected ( Fig 4B). Notably, the subset of DPPA2 sites that lose H3K4me3 are enriched 300 for full-length LINE1 elements and promoters ( Fig 4B). Moreover, the effect on H3K4me3 is specific, 301 since there is no significant change of H3K9me3 in mutants, whilst H3K27me3 is reduced at some loci 302 such as Txnb, but relatively unaffected at most (Fig 4C-D  hypermethylation. In support of this, the differentially methylated promoters (DMP) and LINE1 (DML) 308 correspond to the loci that exhibit reduced H3K4me3 (Fig 4C-D). Furthermore, by investigating all 309 promoters and LINE1, we observe a striking negative correlation (p<2.2e-16) between the level of 310 H3K4me3 loss and of DNAme gain upon Dppa2 KO ( Fig 4F). 311 312 In summary, abrogation of Dppa2/4 is linked with depletion in H3K4me3 at a specific subset of 313 DPPA2-target loci, which directly correlates with acquisition of aberrant DNA hypermethylation.

314
Dppa2/4 could therefore integrate chromatin states to safeguard the pluripotent epigenome, particularly 315 at developmental-associated genes and LINE1 elements. EpiLC leads to a more divergent transcriptome as judged by principal component analysis (Fig 5C), 332 with 801 and 611 differential genes, again preferentially downregulated. Significantly, gene ontology 333 indicated these DEG in EpiLC specifically relate to developmental processes (single-multicellular 334 organism process FDR=0.000004, cell differentiation FDR=0.00012) (Fig S5B), which reflects a 335 general failure to activate genes involved in lineage-specific functions, particularly mesendoderm 336 regulators. For example, Hand 1, Cldn9, Tnxb and others all fail to initiate primed expression in mutant 337 EpiLC (Fig 5A & S5C). This could be linked with the ectopic promoter DNA methylation acquired in 338 the preceeding ESC state. Indeed, the DMP geneset collectively (n=354), which comprises many of the 339 same mesendoderm genes including Hand1, Tnxb, Ttl9, Cldn9 and Gnmt, is significantly upregulated 340 in WT EpiLC (p=0.018) consistent with priming developmental genes, but strikingly, fails to initiate 341 activation in either Dppa2 -/-(p=0.29) or Dppa4 -/-EpiLC (p=0.40), suggesting they have lost 342 competence for expression ( Fig 5D). 343 344 To understand if the impaired expression of mesendoderm genes in EpiLC represents a delay in their 345 activation or stable silencing, we induced endoderm differentiation for 12 days. Dppa2 -/cells appeared 346 morphologically equivalent to WT and activated master endoderm regulators such as Emb and Foxa1 347 with comparable dynamics, indicating no general impairment in differentiation ( Fig 5E). However, 348 endoderm-associated genes including Gnmt, Nkx2-5, Col16a1 and Hand1 all exhibited a highly 349 significant failure to activate in mutants, even after 12 days of endoderm induction, implying an 350 absolute blockade in their response ( Fig 5F). Importantly, Dppa2 is rapidly downregulated after 3 days 351 of endoderm differentiation, but impaired gene upregulation manifests at later timepoints, suggesting a 352 memory of DPPA2 prior activity ( Fig 5G). Indeed, pyrosequencing revealed ectopic promoter DNA 353 methylation established in ESC propagates through to d12 endoderm ( Fig 5H). Together this indicates 354 that the absence of Dppa2/4 in pluripotent phases leads to impaired competence for gene activation 355 during later differentiation. Importantly, cell fate transition per se appears unperturbed, but rather 356 specific genes within the developmental programme are rendered stably epigenetically silenced.

358
We next asked whether repetitive element activation is also affected by Dppa2/4, since many 359 evolutionary young LINE1 become hypermethylated and lose H3K4me3 in their absence ( H3K4me3 and DNAme interact to confer functional epigenetic memory 370 We finally investigated whether induced DNA methylation and H3K4me3 ESC loss is functionally 371 instructive for the subsequent gene silencing memory. We noted that depletion of promoter H3K4me3 Dnmt1 partially rescues their activation block in EpiLC. This effect is significant (p=0.024) among 380 genes with CpG island (CGI) promoters, but not non-CGI promoters (p=0.23) ( Fig 6A). Moreover, we 381 observed re-activation of L1Md_T elements in compound Dnmt1 -/-, Dppa2 -/-ESC and EpiLC (Fig 6B). 382 These data imply that ectopic DNAme in Dppa2 -/cells is instructive, at least at some CpG-dense genes 383 and LINE1, and directly impairs their response to inductive activating signals.

Supplementary Figure 1. Model and enhanced ratiometric reporter for developmental DNA demethylation. A.)
Schematic for design and optimisation of the ratiometric enhanced eRGM cellular DNAme reporter. The system consists of a methylation-sensitive imprinted promoter, which controls expression of GFP according to its level of DNA methylation. The DNA methylation level is set by an antisense upstream genomic region (DNAme sensor) that acquires a similar level of DNAme as the global DNA methylation state, and subsequently adjusts DNAme at the imprinted promoter to equivalence via proximity. This two-stage system generates a robust read-out of global DNAme status (upper panels). Changing the DNAme sensor to a region that is resistant to DNA demethylation (e.g. IAP) prevents eRGM activation in hypomethylated conditions (left FACS plot), confirming that the DNAme sensor status controls activity. Moreover switching the imprinted promoter from Snrpn to Kcnq1ot1 enables a greater degree of expression upon DNA hypomethylation, thereby increasing the dynamic range of the reporter (right FACS plot). Shown in grey is activity in the 'off' hypermethyalted state. Finally, by coupling eRGM with a second methylation-insensitive reporter (Ef1a-mCherry), a single-cell ratiometric score can be generated that normalises for confounding factors. B.) Transcriptomics from ESC maintained in serum/Lif (hypermethylated), titrated t2i/L (hypermethylated) or 2i/L (hypomethylated). The t2i/L and 2i/L transcriptomes are highly comparable despite distinct global methylation states (hyper-and hypo-methylation, respectively), implying transition between these conditions isolates epigenetic resetting without confounding changes in cell identity. C.) Screen shot of genomic methylation pattern from naïve E3.5 epiblast and naïve ESC demonstrating in vitro resetting in 2i/L establishes a highly comparable methylome as in vivo resetting.

D.)
Representative FACS plots of progressive eRGM (GFP) activation by DNA demethylation during 12 day transition from t2i/L to 2i/L in independent eRGM cell lines (Line #1 and #2). Normalising to mCherry (lower panel) enables a ratiometric single-cell readout (each datapoint is a single cell), which closely tracks global methylation levels (right panel). E.) CRISPR screen for gene KO that enable eRGM activation even under hypermethylated conditions (t2i/L) identifies key known DNA methylation and chromatin regulators, confirming eRGM specificity and sensitivity to modulation of epigenetic systems.  values for candidates from independent screens of independent cell lines are highly correlated. FDR<0.05 was used as a threshold. B.) Gene ontology (GO) analysis of candidate factors from the screen shows enrichment for nuclear activity, and involvement in chromatin and/or nucleic acid processes consistent with being epigenetic regulators. C.) Percentage of cells that remain eRGM-negative in hypomethylation-inducing 2i/L upon knockout of the indicated candidate reprogramming factor. Shown is data from KO generated independently in eRGM line#2, analogous to Fig 2A in eRGM line#1 D.) Representative flow cytometry histograms demonstrating knockout of most candidates leads to a significant block of eRGM activation amongst single cells, implying altered epigenetic resetting. Shown is percentage single-cells with eRGM-negative 'off' (marked in grey) after 12 days in DNA hypomethylation-inducing 2i/L culture.

Differentially methylated region (DMR)
in Dppa2  at the genomic sensor region used in eRGM (Dazl) in ESC and EpiLC and protects it from de novo methylation.

DPPA2 CnR
All DPP2 sites