PROSER1 Modulates DNA Demethylation through Dual Mechanisms to Prevent Syndromic Developmental Malformations

TET1, TET2, TET3, DNA methylation, TOPD, Development, Neurodevelopmental disorder. 22 23 24 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint

25 The link between DNA methylation and neurodevelopmental disorders is well established. However, 26 how DNA methylation is fine-tuned – ensuring precise gene expression and developmental fidelity – 27 remains poorly understood. PROSER1, a known TET2 interact or, was recently linked to a severe 28 neurodevelopmental disorder. Here, we demonstrate that PROSER1 interacts with all TET enzymes and 29 stabilizes chromatin -bound TET-OGT-PROSER1-DBHS (TOPD) complexes, which regulate DNA 30 demethylation and developmental gene expression. Surprisingly, we find that PROSER1 also 31 sequesters TET enzymes, preventing widespread demethylation and transposable element de-32 repression. Our findings identify PROSER1 as a key factor which both positively and negatively 33 regulates DNA demethylation essential for mammalian neurodevelopment. 34 35

36 DNA methylation is a fundamental epigenetic process that is essential for normal development. 37 Collaborating with other chromatin -based epigenetic mechanisms, it safeguards the genome by 38 silencing transposable elements, restricts imprinted gene expression, and balanc es gene dosage 39 between sexes (Jones 2012) . Furthermore, DNA methylation can in some cases directly influence 40 transcription factor binding and gene expression through epigenetic modification of promoter and 41 enhancer regions (Schübeler 2015; Rasmussen et al. 2015, 2019; Kreibich et al. 2023). The TET family 42 of DNA demethylases consists of three members (TET1, TET2, and TET3). These closely related proteins 43 contain a conserved catalytic domain that can iteratively oxidize 5 -methylcytosine (5mC) to 5 -44 hydroxymethylcytosine (5hmC), 5 -formylcytosine, and 5 -carboxycytosine, and promote DNA 45 demethylation (Rasmussen and Helin 2016). Mice lacking all three TET enzymes are unable to survive 46 beyond the early stages of development due to gastrulation failures (Dai et al. 2016), while embryos 47 lacking only TET3 can progress to the neonatal stage (Gu et al. 2011). Similarly, although mice lacking 48 either TET1 or TET2 develop normally (Dawlaty et al. 2011)(Moran-Crusio et al. 2011; Quivoron et al. 49 2011)(Li et al. 2011) (Ko et al. 2011) , the combined loss of these enzymes causes developmental 50 abnormalities and increased mortality in a proportion of newborn mice (Dawlaty et al. 2013) . The 51 spectrum of developmental defects in knockout mouse lines indicates that TET enzymes have both 52 unique and redundant roles in maintaining developmental processes during early embryonic 53 development. 54 The function of TET enzymes is modulated via protein -protein interactions with a diverse set 55 of binding partners. For instance, TET enzymes interact with O-linked N-acetylglucosamine (O-GlcNAc) 56 transferase (OGT) (Chen et al. 2013; Vella et al. 2013; Deplus et al. 2013) and the SIN3A-HDAC histone 57 deacetylase complex (Zhang et al. 2015; Williams et al. 2011; Flores et al. 2023; Zhu et al. 2018) to 58 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint promote histone O-GlcNAcylation and histone deacetylation, respectively. In addition, TET1 and TET2 59 interact with the two Drosophila behaviour/human splicing (DBHS) proteins Paraspeckle component 1 60 (PSPC1) and Non -POU Domain Containing Octamer Binding (NONO) (Knott et al. 2016, 2022) to 61 modulate expression of endogenous retroviruses and bivalent genes (Guallar et al. 2018; Huang et al. 62 2022; Li et al. 2020). Finally, TET2 has recently been reported to interact with Proline and Serine Rich 63 1 (PROSER1) in the context of UTX and the MLL3/4 branch of COMPASS (Complex of proteins 64 associated with SET1) to modulate H3K4me1 and H3K4me2 levels at UTX binding sites in the human 65 embryonic kidney cell line HEK293 (Wang et al. 2021). The relative importance of these interactions in 66 regulating TET activity and function is an active area of research. 67 Interestingly, recent findings have linked homozygous loss-of-function mutations in PROSER1 68 to a novel developmental disorder . This condition features hypotonia, developmental delays, 69 genitourinary malformations, and craniofacial abnormalities associated with sensorineural hearing 70 loss and strabismus (Salah et al. 2022) . Given the established role of DNA methylation in 71 neurodevelopment, as evidenced by the association of mutations in DNMT1, DNMT3A, DNMT3B, 72 USP7, and TET3 with a wide spectrum of developmental disorders (Nava and Arboleda 2024) , the 73 observed interaction between PROSER1 and TET2 is intriguing. While this interaction suggests a 74 potential mechanism underlying PROSER1 -associated syndromes, the precise pathological 75 consequences of PROSER1 deficiency on DNA methylation, gene expression, and ultimately 76 developmental integrity during early embryogenesis remain to be elucidated. 77 78

AND DISCUSSION 79 Loss of PROSER1 increases preweaning lethality and is associated with developmental disabiliti es 80 and craniofacial abnormalities 81 To establish a direct causal link between PROSER1 loss -of-function gene mutations and 82 neurodevelopmental disorders, w e generated a mouse line with constitutive inactivation of the 83 endogenous Proser1 gene using CRISPR-Cas9 gene editing (Fig. S1A, B, C). Proser1+/- mice were viable 84 and fertile and these mice were, upon backcrossing to C57BL/6J, intercrossed to generate mice with 85 homozygous PROSER1 loss. Analysis of offspring from these breeders demonstrated that P ROSER1 86 knockout results in partially penetrant pre-weaning lethality (Fig. 1A). We did not observe prominent 87 increases in perinatal lethality suggesting that most PROSER1 knockout embryos may be reabsorbed 88 in utero during early gestation. Surviving P ROSER1 knockout animals weighed less upon reaching 89 adulthood (Fig. 1B) and displayed frequent eye abnormalities including microphthalmia, 90 anophthalmia, and cataracts as well as intermittent tremors and failure-to-thrive (Fig. 1C, 1D ) . To 91 further characterise neuroanatomical defects, we performed microcomputed tomography (microCT) 92 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint scanning of skulls from adult animals (8-14 weeks old) . Consistent with reduced overall weight, 93 volumetric analysis revealed reduced cranial bone volume in PROSER1 knockout animals compared to 94 wildtype littermates (Fig. 1E) , while bone density remained unchanged (Fig. 1F) . Of note, the 95 persistence of this phenotype in fully grown animals suggests that loss of P ROSER1 result s in 96 permanent developmental disability rather than developmental delays. Further comparison of skull 97 shape revealed distinct malformations of the maxillary and frontal bones as well as a tendency for 98 rounded heads and a general shortening of the skull (Fig. 1G) . In summary, germline PROSER1 99 deficiency in mice results in pleiotropic developmental abnormalities that resemble the human 100 neurodevelopmental disorder in which PROSER1 is mutated. 101 102 PROSER1 is a pan-TET interactor and participates in chromatin-associated TOPD complexes 103 Our data demonstrate functional homology between mouse and human PROSER1, suggesting that 104 mouse embryonic stem cells (mESCs) and their differentiation can serve as an appropriate and 105 tractable model system to elucidate the mechanistic role of PROSER1 in early development. A previous 106 study reported a protein-protein interaction between PROSER1 and TET2 in HEK293 cells (Wang et al. 107 2021). To determine the conservation of this interaction in mESCs and its potential extension to the 108 entire TET family of enzymes, we raised two specific anti -murine PROSER1 antibodies with epitopes 109 from PROSER1 C - and N -termini, respectively, and performed endogenous IP -MS using PROSER1 110 knockout mESCs as background control (Fig. 2A, S1D, E, F). Analysis of biological triplicate experiments 111 identified significant enrichment of TET1 and TET2 as well the previously identified TET protein 112 interactors OGT, PSPC1 and NONO. These interactions were also observed in protein lysates isolated 113 from mESC-derived embryoid bodies (EBs). Unlike mESCs, EBs express TET3 and indeed we observed 114 robust enrichment of all three TET enzymes upon PROSER1 IP (Fig. 2B). Importantly, by performing 115 TET2 IP -MS in wildtype and TET2 knockout mESCs we could furthermore recover PROSER1, OGT, 116 PSPC1, and NONO interactions, but no detectable interaction with the other TET enzymes (Fig. 2C). 117 These findings indicate that the presence of TET2 in PROSER1 -containing complexes is mutually 118 exclusive with TET1 and TET3, suggesting that PROSER1 forms discrete complexes with each TET 119 protein. Indeed, PROSER1 IP -MS in TET2 knockout cells still robustly enriched TET1 and TET3, 120 demonstrating that their interaction with PROSER1 is not dependent on TET2 (Fig. S2A). 121 Our data suggest the existence of multiprotein complexes involving TET proteins, OGT, and 122 PROSER1, as well as members of the DBHS family, which are hereafter referred to as TOPD (TET-OGT-123 PROSER1-DBHS proteins) complexes (Fig. 2D). The relative abundance of these complexes is likely to 124 be affected by variation in the expression of TET proteins and their interactors in different tissues. As 125 mentioned above, TET1 and TET2 are highly expressed in mESCs, whereas TET3 expression is induced 126 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint in EBs and upon differentiation to neuronal lineages. Similarly, although the DBHS protein PSPC1 127 robustly associates with TET complexes in all cells tested, PSPC1 can form either PSPC1 -PSPC1 128 homodimers or PSPC1-NONO and PSPC1-SFPQ heterodimers, depending on the relative abundance of 129 each protein (Knott et al. 2016, 2022). Of note, estimation of absolute protein abundance in wildtype 130 mESCs show s that TET proteins and PROSER1 are present at similar copy numbers , whereas the 131 abundance of OGT, PSPC1, and NONO are orders of magnitude higher (Fig. 2E). This implies that excess 132 OGT, PSPC1, and NONO are likely to be involved in processes independent of TET proteins and 133 PROSER1. In contrast, most of the cellular pool of PROSER1 may be engaged within TOPD complexes. 134 Interestingly, PROSER1 IP-MS did not result in enrichment of SIN3A or HDAC1/2 (Fig. 1A, 1B and sup. 135 table S3), suggesting that TET interactions with the SIN3A-HDAC deacetylase complex are independent 136 of PROSER1. We also failed to detect interactions with UTX or members of COMPASS, and profiling of 137 histone modifications by quantitative mass spectrometry revealed little or no global changes in H3K4 138 methylation in two independent PROSER1 knockout mESC lines (Fig. S1G). 139 Initial biochemical cell fractionation of mESCs demonstrated that PROSER1 is predominantly a 140 chromatin-associated protein (Fig. S2 B). We therefore performed PROSER1 chromatin 141 immunoprecipitation and sequencing (ChIP -seq) using our anti -murine PROSER1-N antibody to gain 142 further insights into its function in chromatin. To ensure specificity of enriched peaks, ChIP -seq was 143 carried out simultaneously on wildtype and PROSER1 KO mESCs. Analysis of biological replicate 144 experiments revealed 1712 high-confidence PROSER1 binding sites (Fig. S2C). Upon intersection with 145 publicly available ChIP-seq datasets in mESCs, we found that a large majority (>95%) are co-occupied 146 by TET1, TET2, OGT , PSPC1 or combinations of these (Fig. 2F and S2D). We next asked if PROSER1 147 genome co -localisation with each of the complex components is enriched compared to matched 148 control regions. We observed significant enrichment (P<0.0001, Fisher’s exact test) for TET1, TET2, 149 OGT, and PSPC1 co-localisation (Fig. 2G), whereas no enrichment was observed at CTCF binding sites 150 or gene bodies (Fig. S2E , F). Consistent with a role in gene regulation, PROSER1 high- confidence 151 binding sites are associated with both active (H3K27ac and P300) and repressive (H3K27me3 and 152 SUZ12) chromatin domains (Fig. S2G) and overlap regulatory genomic regions such as promoters, CGIs, 153 and active enhancers which are known to be occupied by TET1 and TET2 (Fig. 2H) (Williams et al. 2011; 154 Rasmussen et al. 2019). 155 156 PROSER1 loss disrupts TOPD complexes and alters TET2 genome-wide chromatin binding 157 To understand how loss of PROSER1 affects the stability of TOPD complexes, we immunoprecipitated 158 TET2 in wildtype and PROSER1 knockout mESC cell lines and analysed eluates by western blotting and 159 label-free mass spectrometry (Fig. 3A, S3A). While PROSER1, OGT and PSPC1, were present in TET2 160 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint eluates from wildtype cells, loss of PROSER1 reduced the relative recovery of OGT and abolished TET2-161 PSPC1 interactions altogether. Furthermore, follow-up analysis of biological triplicate IP eluates using 162 quantitative TMT-labeling demonstrated that TET2-OGT interactions were reduced by ~60% upon loss 163 of PROSER1, while quantitative recovery of SIN3A was not affected (Fig. 3B). Of note, we found that 164 the level of TET2 O-GlcNAcylation was unchanged in PROSER1 knockout cells, suggesting that enzyme-165 substrate interactions between TET2 and OGT are preserved in the absence of PROSER1 (Fig. 3A). 166 To investigate the association between PROSER1, TOPD recruitment to chromatin, and the 167 activation status of PROSER1 -bound regions, we then performed ChIP -seq for TET2 as well as the 168 histone marks H3K4me1 and H3K27ac in wildtype and PROSER1 knockout mESCs (Fig. S3B). Differential 169 enrichment analysis within high- confidence PROSER1 binding sites revealed that loss of PROSER1 170 correlates with reduced TET2 chromatin occupancy, while enrichment of H3K4me1 or H3K27ac does 171 not change in response to loss of PROSER1 within the same regions (Fig. 3C, S3C , D, E). We observed 172 that TET2 binding loss is not limited to PROSER1 binding sites within specific regulatory domains but 173 rather is seen across active enhancers, promoters and CpG islands (CGIs) (Fig. 3D). In contrast to the 174 reduction of TET2 binding at PROSER1-bound genomic regions, we also observed a significant number 175 of sites with increased TET2 occupancy upon knockout of PROSER1 (Fig. 3 E, F). These regions do not 176 show evidence of PROSER1 binding in wildtype cells (Fig. 3E, and S3E), or skewing towards increased 177 association with specific genomic regions compared to TET2 binding sites in wildtype cells (Fig. 3G). 178 However, these regions were linked to increases in H3K27ac upon PROSER1 knockout (Fig. 3E, 3H, S3F), 179 possibly via interactions of TET2 with the histone acetyltransferase P300, as reported previously (Zhang 180 et al. 2017) . Of note, we did not observe global differences in P300 -directed histone acetylation via 181 histone mass spectrometry, nor local differences at matched control regions which were not bound by 182 TET2, suggesting that increased H3K27ac deposition is largely restricted to sites associated with 183 increased TET2 binding (Fig. 3H, 3I, and S3G). Collectively, our results demonstrate that PROSER1 is 184 required for the stability of TOPD protein complexes, and that loss of PROSER1 alters the recruitment 185 of TET2 to chromatin. 186 187 PROSER1 knockout unleashes TET catalytic activity and causes widespread DNA demethylation and 188 de-silencing of endogenous retroviruses 189 To determine the effect of PROSER1 loss on DNA methylation, we harvested genomic DNA from two 190 independent PROSER1 knockout mESC lines and quantified global levels of 5hmdC and 5mdC by mass 191 spectrometry. We observed a decrease in global 5m dC levels, as well as slightly elevated levels of 192 genomic 5hmdC - the major product of TET catalytic activity (Fig. 4A). To investigate which regions are 193 affected by increased TET activity, we generated base- resolution DNA methylation profiles using 194 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint Enzymatic Methyl-Sequencing (EM-seq) in wildtype ESCs (WT), PROSER1 knockout (PROSER1 KO), and 195 PROSER1 knockout cells that were engineered to re-express full-length FLAG -tagged P ROSER1 196 (KO+Rescue) (Fig. 4B). Mapping of EM -seq reads to the mouse genome allowed quantification of 197 cytosine modification states at ~17.5 million CpG sites with at least 10X coverage in all three genotypes. 198 Consistent with our mass spectrometry results, we observed a decrease in DNA methylation in 10 kb 199 windows across the entire genome in P ROSER1 KO cells, while reintroduction of P ROSER1 restored 200 methylation to wildtype levels (Fig. 4C). This widespread DNA hypomethylation was also observed 201 when comparing average DNA methylation levels in diverse genomic regions including 202 heterochromatin, gene bodies, active enhancers and non-CGI promoters (Fig. 4D), though it was noted 203 that regions generally depleted of DNA methylation, such as CGIs and bivalent promoters, were 204 unchanged (Fig. S4A). We furthermore found significant DNA hypomethylation at sites associated with 205 increased H3K27ac deposition identified previously (Fig. 4E and S3B). Importantly, expression and 206 protein copy numbers of the major DNA methylation effectors (DNMT1/UHRF1, DNMT3A/B, TET1 and 207 TET2) were largely unchanged (Fig. 4F, S4B). This implies that DNA methylation changes are a direct 208 consequence of altered TET activity in P ROSER1 knockout cells rather than a result of a general 209 disruption of DNA methylation maintenance machinery. 210 Dual pharmacological inhibition of DNA methylation enzymes and histone deacetylases causes 211 DNA hypomethylation and increased histone acetylation - reminiscent of changes observed upon 212 PROSER1 knockout – and synergizes to cause de-silencing of transposable elements (TEs) (Brocks et al. 213 2017; Daskalakis et al. 2018; Goyal et al. 2023; Cusack et al. 2020) . We therefore used a combination 214 of RNA-seq and TEtranscripts - an analysis pipeline designed to handle reads that map to multiple 215 locations in the genome - to assign multi -mapping reads to specific TE families and analy ze their 216 activity. Loss of PROSER1 led to an increase in transcription of multiple families of long terminal repeat 217 (LTR)-containing endogenous retroviral (ERV) elements, such as ERVK, ERVL, and ERVL-MaLR, whereas 218 expression of the non-LTR L1Md retrotransposons was mildly reduced (Fig. 4G). To understand if this 219 deregulation correlate d with loss of PROSER1, we assessed transcript levels of the differentially 220 expressed TE families upon re- expression of PROSER1. Consistent with the observed restoration of 221 DNA methylation levels described above, expression of differentially expressed TE families (Fig. 4H) – 222 as well as expression of individual TE elements identified solely based on uniquely mapped reads (Fig. 223 4I and S4C) – were restored to near -wildtype levels upon reintroduction of full-length P ROSER1. 224 Collectively, our findings demonstrate that PROSER1 safeguards against genome- wide DNA 225 demethylation and aberrant activation of endogenous retroviruses. 226 227 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint Defective recruitment of TET2 to developmental genes upon PROSER1 loss leads to their subsequent 228 dysregulation during neuronal differentiation 229 To determine the direct effect of altered TET2 chromatin binding , we quantified DNA methylation 230 changes in the high -confidence TET2 differentially bound sites (p- adj<0.05, abs(fold change) ≥2) 231 identified in PROSER1 knockout cells (Fig. 3E). In contrast to regions with gained TET2 binding - which 232 mirrored the genome wide DNA hypomethylation - sites with reduced TET2 binding were instead 233 correlated with increased levels of DNA methylation (Fig. 5A). Further analysis identified that nearly a 234 third of sites with reduced TET2 binding exhibited a significant rise in average DNA methylation (p-235 adj<0.05, minimum 3 CpG per site ) (Fig. 5B). To investigate potential effects on gene expression, we 236 identified enriched gene ontology terms in the subset of 866 genes whose regulatory domains (defined 237 in GREAT as “Basal plus extension” (McLean et al. 2010) ) overlapped with sites with reduced TET2 238 binding. This analysis revealed significant associations with early development processes, such as 239 nervous and skeletal system development, and was linked to mouse knockout phenotypes exhibiting 240 craniofacial abnormalities including eye defects (Fig. 5C). In contrast, genes whose regulatory domains 241 overlapped with sites with increased TET2 binding were largely linked to phenotypes associated with 242 abnormal hematopoietic differentiation (Fig. S5A). 243 To directly assess the effect of P ROSER1 knockout during mESC differentiation, we analysed 244 gene expression in self -renewing mESC lines and in cells differentiated for 2 days towards neuronal 245 lineages (N2B27 monolayer differentiation system). Initial inspection of RNA-seq results revealed that 246 all cell lines successfully exited pluripotency and upregulated markers of early neuronal differentiation 247 (Fig. S5B). Consistent with this, principal c omponent analysis (PCA) revealed differentiation state as 248 the main factor driving variation in the samples (PC1: 50.8% and PC2: 42.3%). In addition, we identified 249 a minor component (PC3: 4.1%) that correlated with P ROSER1 expression, and clearly separated 250 samples cultured for 2 days in N2B27 medium (Fig. 5 D). We therefore examined if the reduced TET2 251 chromatin binding we observed in PROSER1 knockout cells could be linked to gene expression changes 252 at this developmental stage. To do this, we focused on a subset of 60 genes that i) had a regulatory 253 domain overlapping sites of reduced TET2 binding in mESCs, and ii) were differentially expressed (p-254 adj<0.05) between wildtype and PROSER1 knockout cells upon 2 days of neuronal differentiation (Fig. 255 5E and sup. table S7). Contingency analysis revealed significant enrichment of this subset of genes 256 (both increased and decreased in knockout) compared to the subset of differentially expressed genes 257 not associated with reduced TET2 binding sites (7.43% vs 3.91% of total, p-value<0.0001, Fisher’s exact 258 test). Hierarchical clustering identified three clusters (C1-C3) with different expression patterns broadly 259 classified as up -, down - o r m i x e d r e g u l a ti o n i n PROSER1 knockout cells, respectively (Fig. 5E) . The 260 observed increases and decreases in transcript levels upon PROSER1 depletion suggest that TOPD 261 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint complexes possess both activating and repressive regulatory capacities, the specific outcome of which 262 is likely influenced by genomic location and developmental stage. Importantly, comparison of Log2 fold 263 changes in PROSER1 KO vs WT , and KO+Rescue vs WT showed that re-introduction of PROSER1 restored 264 expression of many of these genes – particularly in C1 – to near-wildtype levels (Fig. 5F). Genes co-265 bound by PROSER1/TET2 and whose expression correlated with P ROSER1 expression included 266 important developmental regulators such as homeobox genes (e.g. Pitx2, Barx1, and Pax6), epigenetic 267 regulators (e.g. Cbx4, Cbx8, and Prdm8), and transcription factors (e.g. Tbx15 and Mef2c) implying that 268 PROSER1 plays a regulatory role in their function during early embryonic development (Fig. S5C, D). 269 We furthermore noted significant deregulation of Eyes absent homolog 4 ( Eya4) – a PROSER1/TET2 270 co-bound target gene – that is crucial for eye, heart and sensorineural development (Fig. S5C , D ) 271 (Tadjuidje and Hegde 2013). Collectively, deregulation of these genes may underlie some, if not all, of 272 the neurodevelopmental defects observed upon loss of PROSER1 in mice and humans. 273 In summary, our results are consistent with a dual role for PROSER1 in directing DNA 274 demethylation in early development (Fig. 5G). We show that PROSER1 is a pan -TET interactor that 275 promotes the assembly of TOPD protein complexes and facilitate s their recruitment to chromatin in 276 the proximity of important developmental genes. The recruitment of TET proteins – particularly TET2 277 – at PROSER1-bound sites maintains a lowly methylated state at regulatory regions and appropriate 278 expression of adjacent genes during differentiation. Our findings also indicate that TOPD complexes 279 sequester TET proteins away from other regions of the genome. When PROSER1 is depleted , TET 280 proteins bind to additional sites, resulting in widespread DNA demethylation , and therefore global 281 DNA hypomethylation. Interestingly, the regions with increased TET2 binding also displayed increased 282 levels of H3K27 acetylation. This suggests that P300, a histone acetyltransferase previously shown to 283 interact with TET2 (Zhang et al. 2017) , might be recruited alongside TET2 to these same sites. The 284 combined effects of reduced DNA methylation and increased chromatin openness (caused by P300 285 activity) cooperate to perturb the silencing of endogenous retroviral elements, potentially disrupting 286 the expression of adjacent genes during differentiation. 287 The exact m echanism b y w h i c h T O P D c o m p l e x e s c o n t r o l T E T a c ti v i t y a c r o s s t h e g e n o m e 288 remains unclear. One hypothesis is that certain PROSER1 binding sites act as "sinks," attracting TOPD 289 complexes and TET proteins to specific genomic sites and preventing widespread, uncontrolled DNA 290 demethylation by TET enzymes. Another interesting possibility involves the RNA -binding TOPD 291 component PSPC1, which we have found to interact with TET2 in a PROSER1-dependent manner (Fig. 292 3A and S3A). DBHS proteins including PSPC1 can form higher-order oligomers (Fox et al. 2018; Knott 293 et al. 2016, 2022), potentially functioning as RNA- or DNA-tethered condensates that sequester TOPD 294 complexes, further regulating TET activity. Consistent with this, biochemical fractionation experiments 295 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint in mESCs revealed that the ability of TET2 to fractionate with chromatin is largely dependent on both 296 PSPC1 and its RNA binding capacity (Guallar et al. 2018) . Additional work will be needed to address 297 these possibilities in further detail. 298 Our findings align with prior research. Recent work revealed that introduction of a missense 299 mutation into Tet1 in mESCs leads to a partial disruption of TET1-OGT interaction. This results in global 300 DNA hypomethylation, consistent with the mutated TET1 protein detaching from TOPD complexes to 301 trigger widespread DNA demethylation (Hrit et al. 2018). Similarly, a recent study observed global DNA 302 demethylation and TE de-repression after acute deletion of OGT in mESCs (Sepulveda et al. 2024). This 303 strongly suggests that OGT loss disrupts TOPD complex stability, leading to the release of TET proteins 304 and subsequent genome -wide DNA demethylation. Interestingly, the absence of TET enzymes 305 themselves has been linked to widespread DNA demethylation, particularly in heterochromatin 306 (López-Moyado et al. 2019) . This phenomenon was proposed to stem from disruption of a shared 307 protein complex important for both TET and DNMT enzyme recruitment. When TET enzymes are 308 removed, this complex is disrupted, leading to the redistribution of DNMT enzymes across the genome 309 (López-Moyado et al. 2019) . Thus, in addition to the role of PROSER1 in restraining TET activity, it is 310 interesting to speculate whether TOPD complexes may have additional roles – directly or indirectly - in 311 regulating DNMT activity. 312 Pair-wise TET interactions with specific partner proteins like OGT, PSPC1, and NONO have 313 previously been identified (Chen et al. 2013; Vella et al. 2013; Deplus et al. 2013; Guallar et al. 2018; 314 Huang et al. 2022; Li et al. 2020) . However, our results imply that these interactions do not occur in 315 isolation but can combine to form larger multimeric TOPD complexes that have functional roles in 316 chromatin. Interestingly, mutations in several components of TOPD complexes beyond PROSER1 have 317 been linked to neurodevelopmental disorders. These include X -linked variants of OGT (causing 318 Congenital Disorder of Glycosylation (OGT-CDG))(Pravata et al. 2020; Authier et al. 2024) and NONO 319 (causing NONO-associated syndromic disorder)(Mircsof et al. 2015; Reinstein et al. 2016; Roessler et 320 al. 2023) as well as biallelic loss of TET3 (causing Beck-Fahrner syndrome)(Beck et al. 2020; Seyama et 321 al. 2022), all of which display syndromic developmental delay, intellectual disability and craniofacial 322 dysmorphisms similar to features observed upon inactivation of PROSER1. While mutations in TOPD 323 components are likely to have additional pleiotropic effects, some common features across these 324 disorders may stem from shared disruption of TOPD complexes during development. 325 Chromatinopathies represent an expanding category of congenital developmental disorders 326 arising from disruptions in chromatin function and dysregulation of the epigenome (Nava and 327 Arboleda 2024). Mutations within genes encoding critical epigenome regulators, encompassing both 328 core components and accessory proteins such as PROSER1, contribute to this growing list of 329 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint pathologies. In this study, we show that PROSER1 plays a central role in the assembly of multi-protein 330 chromatin-associated TET complexes that shape the DNA methylome and support gene expression. 331 Furthermore, mice lacking PROSER1 mirror the developmental defects seen in humans with 332 homozygous PROSER1 loss-of-function mutations . We therefore propose that developmental 333 syndromes caused by PROSER1 mutations should be designated as a novel form of chromatinopathy 334 and that f uture investigations into TOPD -related developmental syndromes should leverage the 335 growing understanding of TET enzyme function in development and disease. Our development of a 336 PROSER1 knockout mouse model serves both as a system to gain insight into the underlying 337 mechanisms, and a preclinical model to explore the potential for therapeutic intervention in PROSER1-338 related developmental syndromes. 339 340

341 See supplementary methods for further information. 342 343 DATA AVAILIBILITY 344 Raw and processed data sets are available for download at the Gene Expression Omnibus (GEO) 345 database under the accession number GSE273517. 346 347 COMPETING INTEREST STATEMENT 348 The authors declare no competing interests 349 350

351 The authors thank members of the Rasmussen lab, G. Saredi, and T. Owen -Hughes for advice and 352 discussion. We thank A. Rennie and R. Clarke in the Flow Cytometry unit, C. Gillian in the Biological 353 Resource Unit, A. Atrih and C. Rogers in Fingerprints proteomics facility at University of Dundee as well 354 as A. Tavares and C. Corral in the imaging facility at the University of Edinburgh for technical assistance. 355 Work in the Rasmussen lab was funded by a Cancer Research UK fellowship (C66224/A27092) and 356 through support by University of Dundee. X.L. was funded through a personal scholarship provided by 357 the China Scholarship Council (CSC). Work in the Hajkova laboratory is supported by MRC funding 815 358 (MC_US_A652_5PY70) and an ERC grant (ERC- CoG-648879–dynamicmodifications). The Sidoli lab 359 gratefully acknowledges, for funding , the Hevolution Foundation (AFAR), the Einstein -Mount Sinai 360 Diabetes center, and the NIH Office of the Director (S10OD030286). The authors thank the dedicated 361 team behind the European Galaxy server (UseGalaxy.eu), supported by the German Federal Ministry 362 of Education and Research grant 031L0101C and de.NBI-epi. 363 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint 364 AUTHOR CONTRIBUTIONS 365 A.F and K.D.R. conceived the study. A.F., E.V.K., X.L., and K.D.R. carried out experiments. A.Z., and D.O.C. 366 generated the PROSER1 knockout mouse model. Z.H. and P .H. conducted 5h mdC/5mdC mass 367 spectrometry experiments and analysis. S.S. and S.S. conducted histone mod mass spectrometry 368 experiments and analysis. A.J.B. analyzed mass spectrometry data. A.F. and K.D.R. performed the 369 bioinformatics analysis. A.F. and K.D.R. wrote the original draft of the manuscript followed by review 370 and editing by all authors. K.D.R. supervized the study and acquired funding. 371 372

373 Authier F, Ondruskova N, Ferenbach AT, McNeilly AD, van Aalten DMF. 2024. Neurodevelopmental defects in a mouse 374 model of O-GlcNAc transferase intellectual disability. DMM Dis Model Mech 17: 1–9. 375 Beck DB, Petracovici A, He C, Moore HW, Louie RJ, Ansar M, Douzgou S, Sithambaram S, Cottrell T, Santos-Cortez RLP, et al. 376 2020. Delineation of a Human Mendelian Disorder of the DNA Demethylation Machinery: TET3 Deficiency. Am J Hum 377 Genet 106: 234–245. 378 Brocks D, Schmidt CR, Daskalakis M, Jang HS, Shah NM, Li D, Li J, Zhang B, Hou Y, Laudato S, et al. 2017. DNMT and HDAC 379 inhibitors induce cryptic transcription start sites encoded in long terminal repeats. Nat Genet 49: 1052–1060. 380 Chen Q, Chen Y, Bian C, Fujiki R, Yu X. 2013. TET2 promotes histone O-GlcNAcylation during gene transcription. Nature 493: 381 561–564. 382 Cusack M, King HW, Spingardi P, Kessler BM, Klose RJ, Kriaucionis S. 2020. Distinct contributions of DNA methylation and 383 histone acetylation to the genomic occupancy of transcription factors. Genome Res 30: 1393–1406. 384 https://pubmed.ncbi.nlm.nih.gov/32963030/ (Accessed December 21, 2020). 385 Dai H-Q, Wang B-A, Yang L, Chen J-J, Zhu G-C, Sun M-L, Ge H, Wang R, Chapman DL, Tang F, et al. 2016. TET-mediated DNA 386 demethylation controls gastrulation by regulating Lefty–Nodal signalling. Nature 538: 528–532. 387 http://www.nature.com/articles/nature20095 (Accessed May 10, 2019). 388 Daskalakis M, Brocks D, Sheng YH, Islam MS, Ressnerova A, Assenov Y, Milde T, Oehme I, Witt O, Goyal A, et al. 2018. 389 Reactivation of endogenous retroviral elements via treatment with DNMT-and HDAC-inhibitors. Cell Cycle 17: 811–390 822. 391 Dawlaty MM, Breiling A, Le T, Raddatz G, Barrasa MI, Cheng AW, Gao Q, Powell BE, Li Z, Xu M, et al. 2013. Combined 392 Deficiency of Tet1 and Tet2 Causes Epigenetic Abnormalities but Is Compatible with Postnatal Development. Dev Cell 393 24: 310–323. https://www.sciencedirect.com/science/article/pii/S1534580712005898?via%3Dihub (Accessed May 394 31, 2019). 395 Dawlaty MM, Ganz K, Powell BE, Hu YC, Markoulaki S, Cheng AW, Gao Q, Kim J, Choi SW, Page DC, et al. 2011. Tet1 is 396 dispensable for maintaining pluripotency and its loss is compatible with embryonic and postnatal development. Cell 397 Stem Cell 9: 166–175. http://dx.doi.org/10.1016/j.stem.2011.07.010. 398 Deplus R, Delatte B, Schwinn MK, Defrance M, Méndez J, Murphy N, Dawson MA, Volkmar M, Putmans P, Calonne E, et al. 399 2013. TET2 and TET3 regulate GlcNAcylation and H3K4 methylation through OGT and SET1/COMPASS. EMBO J 32: 400 645–655. http://emboj.embopress.org/cgi/doi/10.1038/emboj.2012.357. 401 Flores JC, Sidoli S, Dawlaty MM. 2023. Tet2 regulates Sin3a recruitment at active enhancers in embryonic stem cells. 402 iScience 26: 107170. https://doi.org/10.1016/j.isci.2023.107170. 403 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint Fox AH, Nakagawa S, Hirose T, Bond CS. 2018. Paraspeckles: Where Long Noncoding RNA Meets Phase Separation. Trends 404 Biochem Sci 43: 124–135. 405 Goyal A, Bauer J, Hey J, Papageorgiou DN, Stepanova E, Daskalakis M, Scheid J, Dubbelaar M, Klimovich B, Schwarz D, et al. 406 2023. DNMT and HDAC inhibition induces immunogenic neoantigens from human endogenous retroviral element-407 derived transcripts. Nat Commun 14. 408 Gu TP, Guo F, Yang H, Wu HP, Xu GF, Liu W, Xie ZG, Shi L, He X, Jin SG, et al. 2011. The role of Tet3 DNA dioxygenase in 409 epigenetic reprogramming by oocytes. Nature 477: 606–612. 410 Guallar D, Bi X, Pardavila JA, Huang X, Saenz C, Shi X, Zhou H, Faiola F, Ding J, Haruehanroengra P, et al. 2018. RNA-411 dependent chromatin targeting of TET2 for endogenous retrovirus control in pluripotent stem cells. Nat Genet 30: 412 733. http://www.nature.com/articles/s41588-018-0060-9. 413 Højfeldt JW, Laugesen A, Willumsen BM, Damhofer H, Hedehus L, Tvardovskiy A, Mohammad F, Jensen ON, Helin K. 2018. 414 Accurate H3K27 methylation can be established de novo by SUZ12-directed PRC2. Nat Struct & Mol Biol 25: 415 225–232. http://www.nature.com/articles/s41594-018-0036-6. 416 Hrit J, Goodrich L, Li C, Wang BA, Nie J, Cui X, Martin EA, Simental E, Fernandez J, Liu MY, et al. 2018. OGT binds a 417 conserved C-terminal domain of TET1 to regulate TET1 activity and function in development. Elife 7. 418 Huang X, Bashkenova N, Hong Y, Lyu C, Guallar D, Hu Z, Malik V, Li D, Wang H, Shen X, et al. 2022. A TET1-PSPC1-Neat1 419 molecular axis modulates PRC2 functions in controlling stem cell bivalency. Cell Rep 39: 110928. 420 https://doi.org/10.1016/j.celrep.2022.110928. 421 Jones PA. 2012. Functions of DNA methylation: islands, start sites, gene bodies and beyond. Nat Rev Genet 13: 484–492. 422 http://www.nature.com/articles/nrg3230. 423 Knott GJ, Bond CS, Fox AH. 2016. The DBHS proteins SFPQ, NONO and PSPC1: A multipurpose molecular scaffold. Nucleic 424 Acids Res 44: 3989–4004. 425 Knott GJ, Chong YS, Passon DM, Liang XH, Deplazes E, Conte MR, Marshall AC, Lee M, Fox AH, Bond CS. 2022. Structural 426 basis of dimerization and nucleic acid binding of human DBHS proteins NONO and PSPC1. Nucleic Acids Res 50: 522–427 535. 428 Ko M, Bandukwala HS, An J, Lamperti ED, Thompson EC, Hastie R, Tsangaratou A, Rajewsky K, Koralov SB, Rao A. 2011. Ten-429 eleven-translocation 2 (TET2) negatively regulates homeostasis and differentiation of hematopoietic stem cells in 430 mice. Proc Natl Acad Sci U S A 108: 14566–14571. 431 Kreibich E, Kleinendorst R, Barzaghi G, Kaspar S, Krebs AR. 2023. Single-molecule footprinting identifies context-dependent 432 regulation of enhancers by DNA methylation. Mol Cell 83: 787-802.e9. 433 https://doi.org/10.1016/j.molcel.2023.01.017. 434 Lerdrup M, Johansen JV, Agrawal-Singh S, Hansen K. 2016. An interactive environment for agile analysis and visualization of 435 ChIP-sequencing data. Nat Struct & Mol Biol 23: 349–357. http://www.nature.com/articles/nsmb.3180. 436 Li W, Karwacki-Neisius V, Ma C, Tan L, Shi Y, Wu F, Shi YG. 2020. Nono deficiency compromises TET1 chromatin association 437 and impedes neuronal differentiation of mouse embryonic stem cells. Nucleic Acids Res 48: 4827–4838. 438 Li Z, Cai X, Cai C-L, Wang J, Zhang W, Petersen BE, Yang F-C, Xu M, Dc W, Yang F-C. 2011. Deletion of Tet2 in mice leads to 439 dysregulated hematopoietic stem cells and subsequent development of myeloid malignancies. Blood 118: 4509–440 4518. http://www.bloodjournal.org/content/118/17/4509.abstract. 441 López-Moyado IF, Tsagaratou A, Yuita H, Seo H, Delatte B, Heinz S, Benner C, Rao A. 2019. Paradoxical association of TET 442 loss of function with genome-wide DNA hypomethylation. Proc Natl Acad Sci U S A 116: 16933–16942. 443 McLean CY, Bristor D, Hiller M, Clarke SL, Schaar BT, Lowe CB, Wenger AM, Bejerano G. 2010. GREAT improves functional 444 interpretation of cis-regulatory regions. Nat Biotechnol 28: 495–501. http://www.nature.com/articles/nbt.1630. 445 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint Mircsof D, Langouët M, Rio M, Moutton S, Siquier-Pernet K, Bole-Feysot C, Cagnard N, Nitschke P, Gaspar L, Žnidariɥ M, et 446 al. 2015. Mutations in NONO lead to syndromic intellectual disability and inhibitory synaptic defects. Nat Neurosci 447 18: 1731–1736. 448 Moran-Crusio K, Reavie L, Shih A, Abdel-Wahab O, Ndiaye-Lobry D, Lobry C, Figueroa ME, Vasanthakumar A, Patel J, Zhao X, 449 et al. 2011. Tet2 loss leads to increased hematopoietic stem cell self-renewal and myeloid transformation. Cancer 450 Cell 20: 11–24. http://linkinghub.elsevier.com/retrieve/pii/S1535610811002224. 451 Nava AA, Arboleda VA. 2024. The omics era: a nexus of untapped potential for Mendelian chromatinopathies. Hum Genet 452 143: 475–495. https://doi.org/10.1007/s00439-023-02560-2. 453 Pravata VM, Omelková M, Stavridis MP, Desbiens CM, Stephen HM, Lefeber DJ, Gecz J, Gundogdu M, Õunap K, Joss S, et al. 454 2020. An intellectual disability syndrome with single-nucleotide variants in O-GlcNAc transferase. Eur J Hum Genet 455 28: 706–714. http://dx.doi.org/10.1038/s41431-020-0589-9. 456 Quivoron C, Couronné L, Della Valle V, Lopez CK, Plo I, Wagner-Ballon O, Do Cruzeiro M, Delhommeau F, Arnulf B, Stern M-457 H, et al. 2011. TET2 inactivation results in pleiotropic hematopoietic abnormalities in mouse and is a recurrent event 458 during human lymphomagenesis. Cancer Cell 20: 25–38. 459 http://linkinghub.elsevier.com/retrieve/pii/S153561081100225X. 460 Ramírez F, Dündar F, Diehl S, Grüning BA, Manke T. 2014. DeepTools: A flexible platform for exploring deep-sequencing 461 data. Nucleic Acids Res 42: 187–191. 462 Rasmussen KD, Berest I, Keβler S, Nishimura K, Simón-Carrasco L, Vassiliou GS, Pedersen MT, Christensen J, Zaugg JB, Helin 463 K. 2019. TET2 binding to enhancers facilitates transcription factor recruitment in hematopoietic cells. Genome Res 464 29: 564–575. http://www.ncbi.nlm.nih.gov/pubmed/30796038 (Accessed April 2, 2019). 465 Rasmussen KD, Helin K. 2016. Role of TET enzymes in DNA methylation, development, and cancer. Genes Dev 30: 733–750. 466 http://genesdev.cshlp.org/lookup/doi/10.1101/gad.276568.115. 467 Rasmussen KD, Jia G, Johansen J V, Pedersen MT, Rapin N, Bagger FO, Porse BT, Bernard OA, Christensen J, Helin K. 2015. 468 Loss of TET2 in hematopoietic cells leads to DNA hypermethylation of active enhancers and induction of 469 leukemogenesis. Genes Dev 29: 910–922. http://genesdev.cshlp.org/lookup/doi/10.1101/gad.260174.115. 470 Reinstein E, Tzur S, Cohen R, Bormans C, Behar DM. 2016. Intellectual disability and non-compaction cardiomyopathy with 471 a de novo NONO mutation identified by exome sequencing. Eur J Hum Genet 24: 1635–1638. 472 Roessler F, Beck AE, Susie B, Tobias B, Begtrup A, Biskup S, Caluseriu O, Delanty N, Fröhlich C, Greally MT, et al. 2023. 473 Genetic and phenotypic spectrum in the NONO-associated syndromic disorder. Am J Med Genet Part A 191: 469–474 478. 475 Salah A, Almannai M, Al Ojaimi M, Radefeldt M, Gulati N, Iqbal M, Alawbathani S, Al-Ali R, Beetz C, El-Hattab AW. 2022. A 476 homozygous frame-shift variant in PROSER1 is associated with developmental delay, hypotonia, genitourinary 477 malformations, and distinctive facial features. Clin Genet 101: 565–570. 478 Schübeler D. 2015. Function and information content of DNA methylation. Nature 517: 321–326. 479 http://www.nature.com/doifinder/10.1038/nature14192. 480 Sepulveda H, Li X, Yue X, Carlos Angel J, Arteaga-Vazquez LJ, Brown C, Brunelli M, Jansz N, Puddu F, Scotcher J, et al. 2024. 481 OGT prevents DNA demethylation and suppresses the expression of transposable elements in heterochromatin by 482 restraining TET activity genome-wide. bioRxiv 1–39. 483 http://biorxiv.org/content/early/2024/01/31/2024.01.31.578097.abstract. 484 Seyama R, Tsuchida N, Okada Y, Sakata S, Hamada K, Azuma Y, Hamanaka K, Fujita A, Koshimizu E, Miyatake S, et al. 2022. 485 Two families with TET3-related disorder showing neurodevelopmental delay with craniofacial dysmorphisms. J Hum 486 Genet 67: 157–164. 487 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint Song Y, Liang Z, Zhang J, Hu G, Wang J, Li Y, Guo R, Dong X, Babarinde IA, Ping W, et al. 2022. CTCF functions as an insulator 488 for somatic genes and a chromatin remodeler for pluripotency genes during reprogramming. Cell Rep 39: 110626. 489 https://doi.org/10.1016/j.celrep.2022.110626. 490 Tadjuidje E, Hegde RS. 2013. The Eyes Absent proteins in development and disease. Cell Mol Life Sci 70: 1897–1913. 491 Vella P, Scelfo A, Jammula S, Chiacchiera F, Williams K, Cuomo A, Roberto A, Christensen J, Bonaldi T, Helin K, et al. 2013. 492 Tet proteins connect the O-linked N-acetylglucosamine transferase Ogt to chromatin in embryonic stem cells. Mol 493 Cell 49: 645–656. http://linkinghub.elsevier.com/retrieve/pii/S1097276512010556. 494 Wang SP, Tang Z, Chen CW, Shimada M, Koche RP, Wang LH, Nakadai T, Chramiec A, Krivtsov A V., Armstrong SA, et al. 495 2017. A UTX-MLL4-p300 Transcriptional Regulatory Network Coordinately Shapes Active Enhancer Landscapes for 496 Eliciting Transcription. Mol Cell 67: 308-321.e6. http://dx.doi.org/10.1016/j.molcel.2017.06.028. 497 Wang X, Rosikiewicz W, Sedkov Y, Martinez T, Hansen BS, Schreiner P, Christensen J, Xu B, Pruett-miller SM, Helin K, et al. 498 2021. PROSER1 mediates TET2 O-GlcNAcylation to regulate DNA demethylation on UTX-dependent enhancers and 499 CpG islands. 5: 1–17. 500 Williams K, Christensen J, Pedersen MT, Johansen J V, Cloos PAC, Rappsilber J, Helin K. 2011. TET1 and 501 hydroxymethylcytosine in transcription and DNA methylation fidelity. Nature 473: 343–348. 502 http://eutils.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&id=21490601&retmode=ref&503 ;cmd=prlinks. 504 Wiśniewski JR, Hein MY, Cox J, Mann M. 2014. A “proteomic ruler” for protein copy number and concentration estimation 505 without spike-in standards. Mol Cell Proteomics 13: 3497–3506. 506 Zhang Q, Zhao K, Shen Q, Han Y, Gu Y, Li X, Zhao D, Liu Y, Wang C, Zhang X, et al. 2015. Tet2 is required to resolve 507 inflammation by recruiting Hdac2 to specifically repress IL-6. Nature 525: 389–393. 508 http://www.nature.com/doifinder/10.1038/nature15252. 509 Zhang YW, Wang Z, Xie W, Cai Y, Xia L, Easwaran H, Luo J, Yen R-WWC, Li Y, Baylin SB. 2017. Acetylation Enhances TET2 510 Function in Protecting against Abnormal DNA Methylation during Oxidative Stress. Mol Cell 65: 323–335. 511 http://linkinghub.elsevier.com/retrieve/pii/S1097276516308516. 512 Zhu F, Zhu Q, Ye D, Zhang Q, Yang Y, Guo X, Liu Z, Jiapaer Z, Wan X, Wang G, et al. 2018. Sin3a-Tet1 interaction activates 513 gene transcription and is required for embryonic stem cell pluripotency. Nucleic Acids Res 46: 6026–6040. 514 515 516 517 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint 518 Figure 1 Loss of PROSER1 increases preweaning lethality and is associated with developmental 519 disabilities and craniofacial abnormalities [A] Histogram showing distribution of genotypes at 520 weaning for 476 pups born from heterozygous Proser1 +/- breeder pairs. The graph shows expected 521 Mendelian numbers (Expected) compared to the actual distribution of genotypes (Observed). The 522 number of homozygous PROSER1 knockout pups at weaning are demonstrably lower than anticipated, 523 with less than 50% being recovered. Statistical significance for contingency table was measured by 524 F i s h e r ’ s e x a c t t e s t ( p < 0 . 0 0 0 1 ) . [ B ] B o x p l o t s h o w i n g b o d y w e i g h t i n P R O S E R 1 k n o c k o u t ( K O ) a n d 525 wildtype (WT) littermates upon reaching adulthood (~8 weeks of age). Comparisons are made for 526 males and females separately ( n=8-10). Statistical significance was measured using two -way ANOVA 527 with multiple comparisons (** p<0.01, *** p<0.001). [C] Bar chart showing percentages of gross 528 abnormalities observed in adolescent and adult PROSER1 knockout animals (n=50). Eye defects include 529 microphthalmia, anophthalmia, corneal ulcers, and cataracts. Other craniofacial defects include 530 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint intermittent head tilt and distorted ear (indicative of otitis media) as well as one case of suspected 531 hydrocephalus. [D] Image of a representative PROSER1 KO animal with microphthalmia. The affected 532 eye is indicated by an arrow. [E] Dot and line graph representing volumetric analysis of microcomputed 533 tomography (microCT) scans of PROSER1 KO and wildtype littermate skulls (n=7). Statistical significance 534 was measured by paired two-tailed t test (* p<0.05) [F] Same as E but representing average CT density 535 (Hounsfield Units) as an indication of bone mineral density. ns indicates not significant. [G] Lateral 536 views of representative microCT scans of a 13-week-old male (above) and 14-week-old female (below) 537 PROSER1 KO animal as well as wildtype littermates. The images are rendered as maximum intensity 538 projections (MIPs) from an equal sized volume to enable direct visual comparison between specimens. 539 Arrows indicate malformations of maxilla and frontal bones as well as rounded head shape. Scale bar 540 indicates skull length of wildtype littermates. 541 542 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint 543 Figure 2 PROSER1 is a pan-TET interactor and participates in chromatin-associated TOPD complexes 544 [A] Volcano plot showing protein hits from α-PROSER1-C immunoprecipitation and mass spectrometry 545 (IP-MS) in WT mESCs (n=3 biological replicates). Protein enrichment is compared to parallel IP -MS in 546 PROSER1 KO mESCs. Dotted lines indicate 2 -fold change and p- value 0.05. Proteins of interest are 547 highlighted and identified. See sup. table S3 for full list of enriched proteins . [B] As in A, but carried 548 out in cells differentiated for 2 days to mouse embryoid bodies (mEBs) where TET3 expression is high 549 compared to mESCs. See sup. table S4 for full list of enriched proteins. [C] As in A, but carried out using 550 α-TET2-N antibody and protein enrichment compared to parallel IP -MS in TET2 KO mESCs. See sup. 551 table S6 for full list of enriched proteins. [D] Illustration of the potential composition of TOPD 552 complexes in mESCs and/or mEBs. The DBHS protein dimer is depicted as consisting of PSPC1 and 553 NONO for simplicity but may be variable in vivo. [E] Copy number estimation (Wiśniewski et al. 2014) 554 of TET proteins and TET interactors in mESCs determined by MS (n=3 biological replicates). The dotted 555 line indicates the sum of TET1, TET2, and TET3 copies, error bars represent the mean ±S.D. [F] ChIP -556 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint seq tracks showing a representative region bound by PROSER1, TET2, OGT and PSPC1 in WT mESCs. 557 PROSER1 ChIP-seq in WT and PROSER1 KO mESCs was carried out in this study ; sources of the other 558 datasets are indicated in the figure. Protein coding genes in the region are indicated below the tracks. 559 CGIs are indicated in gray. [G] Percentage of PROSER1-N peaks or matched controls ( matched to the 560 PROSER1-N peakset in number, size, and distance to DNase hypersensitivity sites in mESCs and 561 generated using Easeq (Lerdrup et al. 2016)) that overlap with TET1, TET2, PSPC1 or OGT binding. All 562 peaksets but PROSER1 were generated from publicly available data (Huang et al. 2022; Rasmussen et 563 al. 2019; Vella et al. 2013; Williams et al. 2011) * p<0.0001, two -tailed Fisher’s exact test. [H] 564 Percentage of PROSER1-N, TET1 or TET2 peaks or matched controls that overlap with active enhancers 565 or promoters. 566 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint 567 Figure 3 PROSER1 loss disrupts TOPD complexes and alters TET2 genome -wide chromatin binding 568 [A] Western blot of input and 10X concentration eluates from α-T E T 2 I P o n l y s a t e s f r o m W T a n d 569 PROSER1 KO1 and 2 mESCs. The intensity of OGT in each eluate was normalized to the intensity of 570 TET2 in the same eluate and the ratios in each IP provided beneath the figure . A slight difference in 571 apparent molecular weight was observed between inputs and eluates owing to different buffer 572 compositions. * indicates a non-specific reactive band. [B] Ratios of OGT:TET2 and SIN3A:TET2 in TMT-573 labelled quantitative mass spectrometry analysis of TET2 immunoprecipitate in WT and PROSER1 KO 574 mESCs. TMT reporter intensity was quantified from 3 biological replicates. Error bars represent the 575 mean ±S.D. * p<0.05, unpaired two -tailed t-test with Welch’s correction. [C] Fold change in TET2 576 binding based on TET2 normalized read counts within PROSER1 peaks or matched control regions upon 577 PROSER1 loss. PROSER1 peaks were sorted by Log2(fold change in PROSER1, KO-WT) and divided into 578 equal-sized quartiles (white). Controls (gray) were generated for each quartile. The effect sizes of 579 binding loss compared to matched control regions were measured with Cohen’s d. * d >0.3 (small 580 effect), ** d >0.6 (medium effect), *** d >0.9 (large effect). ns, non -significant. [D] As C, but TET2 581 normalized read counts are shown at PROSER1 sites within different genomic regions (white) as 582 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint specified below the plot, or controls (gray) generated for the set of PROSER1 peaks within each 583 genomic region. [E] Heatmaps and mean values of normalized ChIP-seq signals for TET2, PROSER1, 584 H3K27ac and H3K4me3 centered at high -confidence (p-adj<0.05, abs(fold change≥2)) sites with gain 585 or loss of TET2 binding. Regions are ranked on p-adj values for TET2 binding between PROSER1 KO and 586 WT. Heatmaps were generated using DeepTools software (Ramírez et al. 2014). [F] Fold change in TET2 587 binding based on TET2 normalized read counts within all significant (p-adj<0.05) differential sites with 588 gain or loss of TET2 binding (white) or matched control regions (gray) upon PROSER1 loss. The effect 589 sizes of binding loss compared to matched control regions were measured with Cohen’s d. * d >0.3 590 (small effect), ** d>0.6 (medium effect), *** d>0.9 (large effect). ns, non-significant. [G] Percentage of 591 all TET2 peaks (Rasmussen et al. 2019) or all sites with significant ( p-adj<0.05) gain of TET2 binding 592 that overlap with genomic regions as specified above the plot. [H] As F , but showing fold change in 593 H3K27ac. [I] Global enrichment of H3K27ac in WT and PROSER1 K O1 and 2 mESCs as measur ed by 594 quantitative MS. n=3 biological replicates of PROSER1 KO lines and 6 of WT. Error bars represent the 595 mean ±SD. Statistical significance was measured by unpaired two-tailed t-test with Welch’s correction. 596 597 598 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint 599 Figure 4 PROSER1 loss unleashes TET activity and cause s widespread DNA demethylation and de -600 repression of endogenous retroviruses [A] Quantification of 5h mdC and 5 mdC by LC -MS/MS in 601 genomic DNA harvested from WT, and PROSER1 KO1 and 2 mESCs. Each symbol represents a sample 602 harvested and processed independently ( n=5 biological replicates with each 2 technical replicates) 603 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint from each cell line. Statistical significance was measured by paired two -tailed t test (paired by same 604 day of harvest due to coordinated fluctuations in 5hmC). ns indicates not significant (* p<0.05, ** 605 p<0.01). [B] Western blot of lysates prepared from WT , PROSER1 KO and PROSER1 KO+Rescue mESCs. 606 GAPDH was probed as a loading control. [C] Violin plot showing average DNA methylation in 10 kb tiles 607 for CpG sites covered by minimum 10 EM -seq reads in all samples in WT, PROSER1 KO and PROSER 608 KO+Rescue mESCs. Lines on violin plot represent the median and quartiles. Statistical significance was 609 measured by Brown-Forsythe and Welch ANOVA test (**** p<0.0001). [D] Quantitation trend plots of 610 DNA methylation quantified by EM-seq in heterochromatin, gene bodies, active enhancers and non -611 CGI promoters for CpG sites covered by minimum 10 EM-seq reads in all samples in WT, PROSER1 KO 612 and PROSER1 KO+Rescue mESCs. [E] XY-scatter plot showing modification state of individual CpG sites 613 covered by at least 10 EM -seq reads that overlap 3196 regions of increased H3K27ac deposition 614 identified previously (Fig. S3B) in PROSER1 KO vs WT cells. Significantly differentially methylated CpG 615 sites (p-adj<0.05, and minimum 25% difference) are highlighted in red for PROSER1 KO vs WT (above) 616 and KO+Rescue vs WT (below). [F] Volcano plot showing protein copy numbers determined by whole-617 proteome MS on PROSER1 KO mESCs ( n=3 biological replicates). Protein enrichment is compared to 618 parallel MS in WT mESCs. Dotted lines indicate 2-fold change and p-adj 0.05. Components of the DNA 619 methylation machinery are highlighted and identified. [G] MA-plot showing TEtranscripts differential 620 expression analysis of TEs in WT and PROSER1 KO mESCs. Each dot represents a sub -family of TE 621 elements and the number of individual elements included in the analysis (covered by at least one 622 unique read) is shown in parentheses. Red and dark red dots indicate significantly differentially 623 expressed TE families at thresholds of p-value<0.05 and p-adj<0.05, respectively. [H] Symbol and line 624 plots comparing expression of TE families in PROSER1 KO vs WT and PROSER1 KO+Rescue vs WT. Plots 625 include TE families found to be differentially expressed (p-adj<0.05) between WT and PROSER1 KO in 626 [G] and split to show upregulated (above) and downregulated (below) TE families. Statistical 627 significance was measured by paired two -tailed t test (* p<0.05, **** p<0.0001). [ I] Tracks showing 628 the region surrounding ETnERV3-int_dup123 and 124 (highlighted in blue) which gain TET2 binding and 629 H3K27ac and become DNA hypomethylated and de-silenced in the absence of PROSER1. Data in the 630 top two panels represent pooled EM-seq methylation calls. Data from WT mESCs are shown in black 631 and data from PROSER1 KO and PROSER1 KO+Rescue are overlaid in purple or blue, respectively. Below, 632 tracks represent ChIP -seq coverage in PROSER1 ChIP -seq ( n=2 biological replicates), and TET2 or 633 H3K27ac ChIP-seq (n=3 biological replicates). The last 6 tracks represent coverage of forward (Fw) or 634 reverse (Rv) transcripts identified by RNA-seq (n=2 biological replicates). 635 636 637 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint 638 Figure 5 Defective recruitment of TET2 to developmental genes upon PROSER1 loss leads to their 639 subsequent dysregulation during neuronal differentiation [A] Quantitation trend plots of DNA 640 methylation for CpG sites covered by minimum 10 EM-seq reads in all samples in WT, PROSER1 KO and 641 PROSER1 KO+Rescue mESCs at high -confidence ( p-adj<0.05, Abs(fold change≥ 2)) sites with loss 642 (above) or gain (below) of TET2 binding upon PROSER1 KO. [B] Pie chart (left) and XY-scatter (right) 643 showing average DNA methylation within high -confidence sites with loss of TET2 binding upon 644 PROSER1 KO. Significantly (p-adj<0.05) hypermethylated or hypomethylated sites (min. 3 CpG per site, 645 min. 10 reads per CpG) are indicated. [C] Bar charts showing enriched gene ontology (GO) terms 646 identified by GREAT (McLean et al. 2010) for genes with a regulatory domain overlapping high -647 confidence sites with loss of TET2 binding upon PROSER1 KO. [D] PCA plot of RNA -seq data. PCA was 648 carried out using Z-scores of the top 2500 variable genes in WT, PROSER1 KO1 and PROSER1 KO+Rescue 649 cells in the mESC state (circular markers) or after 2 days of differentiation towards neuronal progenitors 650 (square markers). [E] Heatmap showing Z-scores of genes differentially expressed at day 2 of neuronal 651 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint differentiation between PROSER1 KO and WT (p-adj<0.05) with regulatory domains overlapping high-652 confidence sites with loss of TET2 binding upon PROSER1 KO in mESCs. Hierarchical clustering 653 identified three distinct clusters C1-C3 as indicated. Gene symbols for selected rows in the heatmap 654 are given on the right of the plot. See sup. table S7 for full list of differentially expressed genes 655 associated with loss of TET2 binding. [F] Bar charts showing average Log 2 fold change for each 656 individual cluster defined in [E] for PROSER1 KO vs WT and KO+Rescue vs WT in mESCs and after 2 days 657 of neuronal differentiation respectively. Statistical significance was measured by Brown-Forsythe and 658 Welch ANOVA test (* p<0.05, **** p<0.0001). [G] Schematic illustrating the dual roles of PROSER1 in 659 directing TET function and DNA demethylation. 660 661 662 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint 663 Figure S1 [A] DNA and amino acid sequences of the region surrounding the PROSER1 gRNA binding 664 site in WT and PROSER1KO mice. Green text indicates the start codon, blue the PAM sequence, purple 665 the splice-site consensus sequence. The BstBI restriction site is indicated above the DNA sequence. 666 PROSER1KO mice contains a 7bp deletion that disrupts the reading frame and introduces a premature 667 stop codon . [B] Representative agarose gel showing genotyping of a wildtype (WT) , PROSER1 668 heterozygous (+/-), and knockout (KO) mouse. bp indicates base pairs [C] Western blot of lysates from 669 the bone marrow of a WT or PROSER1KO mouse. GAPDH was probed as a loading control. [D] As in A, 670 but in three clonal PROSER1 KO mESC lines. When an allele has no purple regions this indicates that 671 the splice-site consensus sequence is lost. Premature stop codons are indicated with red asterisks. 672 Both alleles in these three clones either had a frameshift mutation or lost the splice -site consensus 673 sequence. [B] Western blot of lysates from WT, TET2 KO or PROSER1 KO clonal mESCs. GAPDH was 674 probed as a loading control. [C] Illustration of Mus musculus PROSER1 (not to scale), with the binding 675 regions of our PROSER1 -N and PROSER1 -C antibodies indicated below. The single letter amino acid 676 codes and the residue numbers at which the indicated regions begin and end are given above. [D] 677 Global enrichment of H3K4me1, 2 or 3 in WT and PROSER1 KO1 and 2 mESCs as measured by 678 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint quantitative MS. Error bars represent the mean ±SD. Statistical significance was measured by unpaired 679 two-tailed t-test with Welch’s correction. 680 681 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint 682 Figure S2 [A] Volcano plot showing protein hits from α- PROSER1-C immunoprecipitation and mass 683 spectrometry (IP-MS) in TET2 KO mEBs (n=3 biological replicates). Protein enrichment is compared to 684 parallel IP-MS in PROSER1 KO mEBs. Dotted lines indicate 2-fold change and p-value 0.05. Proteins of 685 interest are highlighted and identified. See sup. table S5 for full list of enriched proteins. [B] Western 686 blot of fractionated WT, TET2 KO and PROSER1 KO1 and 2 mESCs. MEK1/2 and histone H3 were run as 687 fractionation controls for the cytoplasmic and chromatin fractions, respectively. * indicates a non -688 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint specific reactive band. [ C] Volcano plot and frequency histogram showing changes in PROSER1 -689 chromatin binding identified by PROSER1-N ChIP-seq in WT and PROSER1 KO1 mESCs ( n=2 biological 690 replicates). Dotted lines indicate 2-fold change and p-adj 0.05. PROSER1-chromatin binding sites where 691 p-adj<0.05 and abs(fold change≥2) are highlighted in purple and their number given. [D] Venn diagram 692 showing overlap between PROSER1, OGT, TET2, PSPC1 and TET1 peaks in mESCs. The number of 693 overlapping peaks in each combination of peaksets is indicated. All peaksets but PROSER1 -N peaks 694 were generated from publicly available data (Huang et al. 2022; Rasmussen et al. 2019; Vella et al. 695 2013; Williams et al. 2011). [E] Percentage of PROSER1-N peaks or matched controls that overlap with 696 CTCF binding. CTCF binding sites were obtained from publicly available data (Song et al. 2022) . * 697 p<0.0001, two-tailed Fisher’s exact test. [F] As E, but overlapping gene bodies. [G] Heatmaps and mean 698 values of normalized ChIP-seq signals for PROSER1 -N and activating ( P300 and H3K27ac) and 699 repressive (H3K27me3 and SUZ12 ) chromatin features centered at 1712 high -confidence PROSER1 700 binding sites in mESCs. P300 and SUZ12 enrichment were generated from (Wang et al. 2017) and 701 (Højfeldt et al. 2018) respectively, and regions were ranked based on the H3K27me3 ChIP-seq signal. 702 Heatmaps were generated using DeepTools software (Ramírez et al. 2014). 703 704 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint 705 Figure S3 [A] Estimated protein abundance indicated by exponentially modified protein abundance 706 index (emP AI) of TET2, OGT , PSPC1 and PROSER1 in TET2 IP-MS normalized to TET2 IP efficiencies in 707 each experiment (n= 1 biological replicate). n.d., not detected. [B] Volcano plot and frequency 708 histograms showing changes in TET2 (left), H3K4me1 (center) or H3K27ac (right) enrichment on 709 chromatin identified by ChIP-seq in WT and PROSER1 KO1 mESCs (n= 3 biological replicates). Dotted 710 lines indicate 2 -fold change and p- adj 0.05. Sites where p-adj<0.05 and Abs(fold change ≥2) are 711 highlighted in blue or black (for TET2 or H3K4me1 and H3K27ac, respectively) and their number is 712 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint given. [C] Fold change in PROSER1 binding (left) or TET2 binding (right) based on normal ized read 713 counts within PROSER1 DOWN and PROSER1 UP sites (p-adj0.3 (small effect), ** d>0.6 (medium effect), *** d>0.9 (large effect). ns, 716 not-significant. [D] Fold change in PROSER1 binding (above) or enrichment of H3K4me1 (center) or 717 H3K27ac (below) within PROSER1 peaks or matched control regions upon PROSER1 loss. PROSER1 718 peaks were sorted by Log 2(fold change in PROSER1, KO- WT) and divided into equal -sized quartiles 719 (white). Controls (gray) were generated for each quartile. The effect sizes of binding loss compared to 720 matched control regions were measured with Cohen’s d. * d >0.3 (small effect), ** d >0.6 (medium 721 effect), *** d>0.9 (large effect). ns, non-significant. [E] Percentage of PROSER1 peaks that overlap with 722 high-confidence (p-adj<0.05, Abs(fold change≥2)) sites with gain or loss of TET2 binding. * p<0.0001, 723 two-tailed Fisher’s exact test. [F] As D, but fold change in TET2 binding ( left) or H3K27ac enrichment 724 (right) at quartiles of all sites with significant (p-adj<0.05) gain of TET2 binding. [G] Global enrichment 725 o f H 3 K 5 a c ( l e ft ) , H 4 K 8 a c ( c e n t e r ) a n d H 3 K 1 8 a c ( r i g h t ) i n W T a n d P R O S E R 1 K O 1 a n d 2 m E S C s a s 726 measured by quantitative MS. n=3 biological replicates of PROSER1 KO lines and 6 of WT. Error bars 727 represent the mean ±SD. Statistical significance was measured by unpaired two -tailed t-test with 728 Welch’s correction. 729 730 731 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint 732 Figure S4 [A] Quantitation trend plots (window size 50 bp, step 50 bp, 1 kb flanking sequence) of DNA 733 methylation quantified by EM-seq in bivalent promoters and CGIs for CpG sites covered by minimum 734 10 EM-seq reads in all samples in WT, PROSER1 KO and PROSER1 KO+Rescue mESCs. [B] Coverage of 735 forward (Fw) or reverse (Rv) transcripts from RNA -seq (n=2 biological replicates) in WT, PROSER1 KO 736 and KO+Rescue mESCs at the regions surrounding Dnmt1 (left) and Uhrf1 (right). [C] Box plots showing 737 TEtranscripts differential expression analysis of TEs. Dots represent expression of individual TEs whose 738 expression was quantified using only uniquely mapped RNA -seq reads. The box plots compare 739 differential expression of TEs in PROSER1 K O vs WT and PROSER1 K O+Rescue vs WT for ET nERV3-int 740 (left), RLTR13B2 (middle), and IAPEy-int (right) families of endogenous retroviral elements, all found to 741 be significantly deregulated upon PROSER1 KO. Statistical significance was measured by unpaired two-742 tailed t-test with Welch’s correction (* p<0.05). 743 744 745 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint 746 Figure S5 [A] Bar charts showing enriched gene ontology (GO) terms identified by GREA T (McLean et 747 al. 2010) for genes with regulatory domain overlapping high-confidence sites with gain of TET2 binding 748 upon PROSER1 KO. [B] Heatmap showing the changes in expression (Z-score of normalized read counts 749 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint from RNA-seq) of key pluripotency markers (above), and neuronal markers (below) in WT, PROSER1 KO 750 and PROSER1 KO+Rescue mESCs (left) or cells subjected to monolayer differentiation in N2B27 medium 751 (right). Gene symbols for each row in the heatmap are given on the right of the plot. [C] As in B, but 752 for selected developmental genes co-bound by PROSER1 and TET2 whose expression correlated with 753 PROSER1 expression. [D] Tracks showing the region surrounding the developmental genes Prdm8, 754 Eya4, Tbx15, and Cbx4. Data in the top two panels represent pooled EM -seq methylation calls. Data 755 from WT mESCs are shown in black and data from PROSER1 KO and PROSER1 KO+Rescue are overlaid 756 in purple or blue, respectively. Below, tracks represent ChIP-seq coverage in PROSER1 ChIP-seq (n=2 757 biological replicates), and TET2 or H3K27ac ChIP-seq (n=3 biological replicates). CGIs are indicated in 758 gray. 759 760 761 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted August 2, 2024. ; https://doi.org/10.1101/2024.07.31.606086doi: bioRxiv preprint