Mammalian DNA maintenance methylation depends on histone but not PAF15 ubiquitination

Abstract

In mammals, DNA methylation is an essential inheritable epigenetic modification maintained by the DNMT1-UHRF1 machinery in a UHRF1 E3 ubiquitin ligase-activity-dependent manner. UHRF1-catalyzed histone ubiquitination and PAF15 ubiquitination have been proposed to recruit DNMT1 to hemi-methylated CpGs. Here, we confirm that both PAF15 and histone H3 are robustly ubiquitinated by UHRF1 in a replication-dependent manner during S phase. However, loss of PAF15 or its ubiquitination, or overexpression of ubiquitination-deficient PAF15, does not alter DNMT1 recruitment to replicating DNA or global DNA methylation. Notably, the DNMT1 inhibitor GSK-3685032 strongly enhances histone ubiquitination, likely via UHRF1 activation driven by accumulated hemi-methylated CpGs, but markedly reduces PAF15 ubiquitination. We show that UHRF1-mediated PAF15 ubiquitination requires DNMT1, is strictly replication-coupled, and that PAF15 and histone ubiquitination are mediated by distinct UHRF1 pools. Taken together, our study demonstrates that hemi-methylated-DNA-stimulated local histone ubiquitination, rather than replication-coupled PAF15 ubiquitination, drives DNMT1 targeting for maintenance methylation.

Here the authors demonstrate that hemi-methylated-DNA-stimulated local histone ubiquitination, rather than replication-coupled PAF15 ubiquitination, can drive DNMT1 targeting for maintenance of DNA methylation.

Introduction

In mammalian cells, methylation of the fifth carbon atom of cytosine within CpG dinucleotide sequences is a highly heritable epigenetic modification with pivotal biological roles in embryonic development, cell differentiation and fate determination, transcriptional regulation, and genome stability¹. In principle, DNA methylation modulates chromatin structure and regulates interactions between trans-acting factors and targeted DNA sequences². DNA maintenance methylation, exerted predominantly by DNMT1 (DNA methyltransferase 1), can faithfully transmit DNA methylation patterns from parental to newly synthesized DNA strands during mitosis and is crucial for maintaining cellular identity and function^3,4.

UHRF1 (ubiquitin-like with PHD and RING finger domains 1) is an essential cofactor for DNMT1-mediated DNA maintenance methylation, guiding the recruitment of DNMT1 to hemi-methylated CpG sites generated during semiconservative DNA replication^5–9. UHRF1 possesses multiple specialized structural and functional domains, namely SRA (binding hemi-methylated DNA^7–9), PHD (interacting with histone H3^10,11), TTD (binding H3K9me2/3^12–14 and LIG1K126me¹⁵), and RING E3 ligase domain that collectively enable its precise localization to newly replicated hemi-methylated DNA. DNMT1 is recruited to nascent DNA in a UHRF1-dependent manner^5,6. Further studies highlight the critical role of UHRF1’s E3 ubiquitin ligase activity in the recruitment of DNMT1 and consequently maintenance of DNA methylation^16,17. Accumulative evidence supports that, upon binding hemi-methylated DNA, UHRF1 undergoes conformational changes that activate its E3 activity, leading to histones H3 and H2B ubiquitination^18–23. DNMT1, via a ubiquitin-interacting motif (UIM) within its replication foci targeting sequence (RFTS) domain, can bind ubiquitinated H3/H2B, which presumably, together with the interaction with UHRF1, leads to DNMT1 recruitment, activation, and subsequent methylation of hemi-methylated DNA to establish fully methylated DNA^{16,17,23–28}.

Recently, Makoto Nakanishi and colleagues reported a PAF15 ubiquitination-dependent alternative mode for DNMT1 recruitment and DNA maintenance methylation²⁹. PAF15 (KIAA0101) was initially identified as a cell cycle-regulated, PCNA-associated factor that plays roles in cell cycle progression and DNA damage response^30–33. A later study identified PAF15 as a major UHRF1 ubiquitination substrate³⁴. Sharing sequence homology with the histone H3 N-terminal tail, PAF15 can undergo dual mono-ubiquitination (PAF15Ub2) in S phase by UHRF1²⁹, and the RFTS domain of DNMT1 also binds double monoubiquitination of PAF15³⁵. Using a Xenopus laevis egg extract replication system, it was shown that UHRF1 ubiquitinates PAF15 in early S phase and histone H3 in late S phase²⁹. Notably, introducing the K/R ubiquitination-deficient mutation into the endogenous Paf15 gene was shown to significantly impair global DNA methylation in mESCs²⁹. It was proposed that H3 ubiquitination serves as a backup for PAF15 ubiquitination-mediated DNMT1 recruitment, and both PAF15 and H3 ubiquitination contribute to high fidelity of DNA methylation inheritance^29,36. This hypothesis is attractive because the PAF15 pathway is independent of histones and could, in theory, allow DNA maintenance methylation to occur right after DNA replication but before chromatin assembly, whereas the histone ubiquitination pathway requires the presence of nucleosomes and thus more likely occurs after chromatin assembly. However, as the previous assessment of the Paf15 K/R ubiquitination-deficient mutation on DNA methylation was carried out with mESCs cultured in 2i condition²⁹, which is known to result in global DNA hypomethylation because DNA maintenance methylation activity is severely compromised^37–40, the exact role and to what extent the PAF15 pathway contributes to DNA maintenance methylation in mammalian cells remain to be evaluated.

In this study, we extensively tested the role of PAF15 in DNA methylation in mammalian cells. Unexpectedly, we found that PAF15 knockout (KO) in multiple cell types has no significant effect on DNA methylation or DNMT1 recruitment. Similarly, ectopic overexpression of a ubiquitination-defective PAF15 mutant also had no significant effect on DNA methylation. Moreover, a Paf15 knock-in mutation that specifically abolishes UHRF1-mediated ubiquitination had no significant effect on DNA methylation in mouse E14 embryonic stem cells (mESCs). Notably, we found that while histone ubiquitination by UHRF1 is strongly activated by hemi-methylated DNA, PAF15 ubiquitination is a DNMT1-promoted, DNA replication-coupled event independent of hemi-methylated DNA. Our study demonstrates that the DNMT1-dependent, UHRF1-mediated PAF15 ubiquitination is essentially dispensable for DNA methylation, thus supporting UHRF1-mediated histone ubiquitination being the key pathway essential for local DNMT1 recruitment and DNA maintenance methylation.

Results

PAF15 ubiquitination occurs throughout the S phase, including the late stage that replicates pericentromeric heterochromatin DNA

To compare the roles of PAF15 and histone ubiquitination in DNA maintenance methylation, we first aimed to define their relationship with DNA replication. We synchronized HeLa and NIH3T3 cells at the G1/S transition point with a dual synchronization strategy using nocodazole and aphidicolin⁴¹. Cells were then released into S phase by culturing in fresh medium lacking nocodazole or aphidicolin, and samples were collected at various time points to assess PAF15 and histone H3 ubiquitination levels along S phase progression by Western blot (WB) analysis, with antibody specificity validated in Supplementary Fig. 1. Consistent with previous reports^29,32,42, in both cell lines PAF15 mono- and di-ubiquitination were detected only after release into S phase, reached a peak level around 3 h and then declined (Fig. 1a, b and Supplementary Fig. 2). On the other hand, WB analysis using an H3 specific antibody revealed that histone H3 mono-ubiquitination level steadily increased and reached the peak 9 h after cell cycle release. WB analysis using an H3K18Ub-specific antibody⁴³ revealed a parallel increase of H3K18Ub up to 7 hours, followed by a reduction at 9 hours (Fig. 1a, b). The discrepancy between H3Ub and H3K18Ub is likely due to the existence of UHRF1-independent H3 ubiquitination in G2/M phase (see below). The assignment of mono- and di-ubiquitinated PAF15 and mono-ubiquitinated H3 (PAF15-endo-Ub, PAF15-di-endo-Ub, and H3-endo-Ub) was supported by the detection of parallel up-shifted bands with expected molecular weights in 293T cells expressing an increasing level of exogenous 6×His-Myc-Ub (PAF15-exo-Ub, PAF15-di-exo-Ub, and H3-exo-Ub) (Fig. 1c). Note that expression of increasing exogenous 6×His-Myc-Ub led to progressively decreased PAF15 and H3 ubiquitination by the endogenous ubiquitin (Fig. 1c), most likely due to the competition of exogenous expressed 6×His-Myc-Ub. To accurately define S phase progression, we made use of EdU (5-ethynyl-2’-deoxyuridine) pulse labeling. The results in Fig. 1d demonstrated that in HeLa cells, DNA synthesis peaked at 3 h and then declined, revealing a robust correlation between DNA replication levels and PAF15 ubiquitination.

Fig. 1. PAF15 is ubiquitinated throughout S phase and associated also with late-replicating pericentromeric heterochromatin.

a WB analysis of PAF15 and histone H3 ubiquitination along S phase progression in HeLa cells. Cells were synchronized at G1/S transition via nocodazole and aphidicolin double-block, released into S phase with fresh medium, and collected at indicated time points. LE, lower exposure. b As in (a), except experiments were performed in NIH3T3 cells. c WB analysis of endogenous and 6×His-Myc-Ub-conjugated PAF15 and H3. 293 T cells were transfected with a control vector or plasmids expressing 6×His-Myc-Ub. Whole-cell extracts were subjected to WB analysis using anti-Ub (left panel), anti-PAF15 (middle panel), and anti-H3 antibody (right panel), respectively. PAF15-endo-Ub, PAF15 mono-ubiquitinated by endogenous Ub; PAF15-exo-Ub, PAF15 mono-ubiquitinated by 6×His-Myc-Ub; PAF15-di-endo-Ub, PAF15 di-ubiquitinated by endogenous Ub; PAF15-di-exo-Ub, PAF15 di-ubiquitinated by 6×His-Myc-Ub; H3-endo-Ub, H3 mono-ubiquitinated by endogenous Ub; H3-exo-Ub, H3 mono-ubiquitinated by 6×His-Myc-Ub. d LC‑MS quantification of dynamic DNA replication in synchronized HeLa cells. Cells were synchronized at G1/S, released into S phase, and pulse‑labeled with 10 μM EdU for 15 min at indicated time points. EdU incorporation was quantified by LC‑MS. Data are presented as mean ± SD from three independent biological replicates. e Representative immunostaining images showing PAF15 colocalization with EdU-labeled newly replicated DNA along S phase progression in HeLa cells. Intensity profiles of fluorescence at marked positions are shown below corresponding images. Scale bar, 20 μm. f As in (e), except experiments were performed in NIH3T3 cells. g WB analysis of PAF15 and histone H3 ubiquitination in wild-type and Uhrf1-KO mouse E14 cells. h WB analysis of PAF15 and histone H3 ubiquitination in wild-type and UHRF1-KO HeLa cells at different cell cycle stages. For knockout of UHRF1 in HeLa cells, cells were transfected with the lentiCRISPRv2 vector or sg-UHRF1 and then selected with 4 μg/mL puromycin for three days. Cells were synchronized to the G1/S boundary by aphidicolin treatment, S phase by releasing from G1/S for 4 hours, and to the G2/M phase by nocodazole. Ctr, unsynchronized cells. WB results in (a–c, g, h) were consistent across two independent biological replicates.

We also monitored the dynamic subcellular localization of PAF15 during S phase. In HeLa cells, 15 min EdU pulse labeled euchromatic DNA in early S phase (1–3 hours) and late-replicating heterochromatin puncta in late S phase (5–7 hours). PAF15 colocalized nicely with EdU-labeled DNA, with a diffuse pattern at 3 hours and a focal pattern at 5–7 hours (Fig. 1e and Supplementary Fig. 2a, c, d). Unlike HeLa cells, mouse NIH3T3 fibroblast cells have pericentromeric heterochromatin that forms typical bright DAPI foci. EdU pulse-labeling assay revealed that in NIH3T3 cells, PAF15 displayed a diffuse pattern at the early stage of DNA replication (1 hour) and then a clear focal pattern that overlapped nicely with the late-replicating heterochromatin (3 hours and 5 hours) (Fig. 1f and Supplementary Fig. 2b, c, d). Thus, PAF15 exhibits a ubiquitination pattern and dynamic localization that is associated with the entire DNA replication process, not just the early-replicating DNA. This observation is consistent with the identification of PAF15 being a PCNA-interacting protein with a role in DNA replication^30,31.

To confirm whether UHRF1 is responsible for both PAF15 and histone H3 ubiquitination during the S phase^16,29,34, we compared the ubiquitination levels of both proteins in wild-type and Uhrf1-KO mESCs. As shown in Fig. 1g, loss of UHRF1 protein markedly reduced the ubiquitination levels of both PAF15 and histone H3 in mESCs. Likewise, UHRF1 knockout in HeLa cells also drastically reduced the ubiquitination levels of PAF15 and histone H3 in unsynchronized cells (Fig. 1h, compare lane 5 with lane 1). We also synchronized the control and UHRF1-KO HeLa cells to G1/S, S (release from G1/S for 4 hours), and G2/M phases and examined the levels of PAF15 and H3 ubiquitination by WB analysis (Fig. 1h and Supplementary Fig. 2e, f). Notably, UHRF1 knockout impaired both PAF15 and H3 ubiquitination in S phase, thus confirming that UHRF1 is responsible for both PAF15 and histone H3 ubiquitination in S phase. It is noteworthy that loss of UHRF1 did not impair H3 ubiquitination levels in G2/M phase, as detected by WB using an anti-H3 antibody. The H3 ubiquitination in G2/M phase is not catalyzed by UHRF1, but by G2E3, an E3 ligase that specifically catalyzes H3 ubiquitination at the K14 site⁴⁴. In support of this notion, WB analysis using an H3K18Ub-specific antibody detected only H3K18Ub in control S phase cells, consistent with the previous report that UHRF1 can catalyze histone ubiquitination at K18 and K23 sites^17,19. These data also explain why total H3 ubiquitination increases progressively along the cell cycle until G2/M, whereas PAF15 ubiquitination parallels S phase.

Collectively, our data indicate that UHRF1-mediated PAF15 ubiquitination occurs throughout S phase and correlates closely with DNA replication activity. On the other hand, E3 ligase(s) other than UHRF1 also contribute to histone H3 ubiquitination, especially in G2/M phase⁴⁴, which together result in a progressively increased H3 ubiquitination along S to G2/M phase of the cell cycle.

PAF15 is not essential for global DNA methylation in HeLa and 293 T cells

It was shown that in Xenopus egg extracts, PAF15 ubiquitination facilitates DNMT1 recruitment and supports DNA methylation maintenance^29,36. In addition, introducing the K/R ubiquitination-deficient mutation into the endogenous Paf15 gene was shown to significantly impair global DNA methylation in naive mESCs, particularly in early-replicating and highly methylated domains (HMDs)²⁹. However, whether PAF15 is a general key component of mammalian DNA maintenance methylation machinery has not been addressed.

To address the role of PAF15 in DNA maintenance methylation, we made use of CRISPR-Cas9 technology to knock out PAF15 in both HeLa and 293 T cells (Supplementary Fig. 3a). The pools of PAF15-KO HeLa cells were obtained, cultured for two weeks, and the efficient PAF15 knockout was confirmed via loss of PAF15 protein by WB analysis, effective cleavage of target sequences by T7E1 assay and by DNA sequencing (Fig. 2a and Supplementary Fig. 3b). Genomic DNA was then prepared from both control and PAF15-KO cells, and global DNA methylation levels were quantitatively measured by high-performance liquid chromatography (HPLC) and liquid chromatography-mass spectrometry (LC-MS). To our surprise, these analyses revealed that PAF15 knockout had no significant effect on global DNA methylation levels in HeLa cells (Fig. 2b, c). Similarly, PAF15 knockout in 293 T cells also had negligible effects on global DNA methylation levels (Fig. 2d, e and Supplementary Fig. 3c). Amplicon-based bisulfite sequencing of repetitive elements LINE1 and HERV-K further showed that PAF15 knockout did not have significant effect on DNA methylation in these highly methylated regions (Fig. 2f, g).

Fig. 2. PAF15 is dispensable for DNA methylation in human cells.

a WB analysis of PAF15 protein levels (upper panel) and T7 endonuclease I validation of genome editing efficiency (lower panel) in vector control (sg-vec) and PAF15-KO (sg-PAF15) HeLa cells. Three different sgRNAs all led to efficient PAF15 gene editing. b The levels of genomic DNA methylation (mC) in vector control and PAF15-KO HeLa cells as in (a) were determined by HPLC. Data are presented as mean ± SD from three independent biological replicates. N/C, no significant change (relative difference <5%). c As in (b), except experiments were performed by LC-MS. d WB analysis of PAF15 expression (upper panel) and T7 endonuclease I-based assessment of PAF15 gene editing (lower panel) in vector control (sg-vec) and PAF15-KO (sg-PAF15) 293 T cells. PAF15 knockout was carried out by co-transfection of three sgRNAs. e The levels of genomic DNA methylation (mC) in vector control and PAF15-KO 293 T cells determined by HPLC. Data are presented as mean ± SD from three independent biological replicates. N/C, no significant change (relative difference <5%). f The levels of DNA methylation for each indicated CpG site in LINE1 and HERV-K in vector control and PAF15-KO HeLa determined by amplicon bisulfite sequencing. g As in (f), experiments were performed in 293 T cells. h WGBS analysis of global CpG methylation levels in vector control and PAF15-KO HeLa cells using one independent biological sample. i Bimodal distribution of CpG methylation levels across the genome from the same WGBS dataset in (h). j CpG methylation levels at CpG islands (CpGi), gene bodies, promoters and transposable-element (TE) regions from the same WGBS dataset in (h). k DNA methylation levels within early- and late-replicating genomic regions from the same WGBS dataset in (h). l Representative genome-browser tracks illustrating DNA methylation profiles from the same WGBS dataset in (h). For box plots in (h, j, and k): lines, median; boxes, 25th–75th percentiles; whiskers, 1.5 × interquartile range from the box. P values from one-way ANOVA with Tukey’s post hoc test.

To rule out potential compensation by the de novo methyltransferases DNMT3A and DNMT3B, we performed PAF15 knockout in DNMT3A/B double-knockout (DNMT3A/B-DKO) HeLa cells and monitored the effect on global DNA methylation over time. We used two independent DNMT3A/B-DKO HeLa clones (#15 and #17), both confirmed to lack DNMT3A and DNMT3B expression (Supplementary Fig. 3d). Global DNA methylation levels determined by quantitative HPLC revealed a ~8.8% reduction in clone #15 and ~5.9% reduction in clone #17 relative to wild-type HeLa cells (Supplementary Fig. 3e). However, knockout of PAF15 in both DNMT3A/B-DKO clones had no further reduction of global DNA methylation measured at day 5 and day 19 post-transfection of corresponding sg-PAF15 plasmids, and the effective PAF15 knockout was confirmed by WB analysis of the same set of samples (Supplementary Fig. 3f, g). To further confirm this result, we isolated two single-cell-derived PAF15-KO clones from the #17 background and validated the lack of PAF15 protein by WB (Supplementary Fig. 3h). Further analysis revealed no significant reduction in global DNA methylation in these clones (Supplementary Fig. 3i). Collectively, these data indicate that, at least in HeLa cells, PAF15 has no significant contribution to DNA methylation even in the absence of DNMT3A/B.

To further and more accurately assess the effect of PAF15 knockout on DNA methylation, we analyzed DNA methylation status at base-pair resolution by whole-genome bisulfite sequencing (WGBS). Wild-type and PAF15-KO HeLa cells showed comparable CpG methylation levels (60.69% vs. 58.71%) (Fig. 2h). Both cell lines exhibited bimodal methylation distributions without notable differences (Fig. 2i), and metagene analysis of methylation levels across various genomic elements revealed no significant changes in DNA methylation (Fig. 2j). Given that PAF15 ubiquitination is proposed to promote DNA methylation in early-replicating regions in mESCs²⁹, we also evaluated the effect of PAF15 knockout on DNA methylation of early-replicating DNA using published data for HeLa cells⁴⁵. This analysis revealed that, compared to the late-replicating DNA, PAF15 knockout did have a very marginal effect on DNA methylation in early-replicating regions (Fig. 2k). As an example, visualization of methylation profiles at a chromosome segment containing both early- and late-replicating regions revealed no substantial changes in the overall DNA methylation landscape upon loss of PAF15 (Fig. 2l).

PAF15 knockout does not alter global DNA methylation in NIH3T3 and mESCs

To assess the role of PAF15 in regulating DNA methylation in non-cancerous cell lines and its functional conservation across species, we also employed the CRISPR-Cas9 technology to knockout Paf15 in NIH3T3 cells and mESCs (Supplementary Fig. 4a). In NIH3T3 cells, all three gRNAs successfully knocked out Paf15 (Fig. 3a and Supplementary Fig. 4b). HPLC and amplicon-based bisulfite sequencing analyses of LINE1 revealed that Paf15 disruption did not significantly alter global DNA methylation levels or the methylation status of LINE1 repetitive sequences in NIH3T3 cells (Fig. 3b, c). To rule out the influence of cell population heterogeneity, we also obtained a single-cell-derived Paf15-KO NIH3T3 cell line. The successful knockout of Paf15 in this monoclonal population was confirmed via both DNA sequencing and WB analysis (Fig. 3d and Supplementary Fig. 4c). Consistent with the results with the pooled knockout populations, no significant difference in global DNA methylation was observed between control cells and monoclonal Paf15-KO NIH3T3 cells (Fig. 3e). Moreover, we generated a population of Paf15-KO mESCs by using both sg-Paf15−2 and sg-Paf15−3 (Fig. 3f and Supplementary Fig. 4d). Measurement of DNA methylation level by HPLC analysis and immunostaining with an anti-5mC antibody demonstrated that the global DNA methylation levels in Paf15-KO mESCs were comparable to those in control cells (Fig. 3g, h). As a positive control, Uhrf1 knockout in mESCs markedly reduced the global DNA methylation levels (Fig. 3g, h).

Fig. 3. PAF15 is dispensable for DNA methylation in mouse cells.

a WB analysis of Paf15 expression (upper panel) and T7 endonuclease I-based assessment of Paf15 gene editing (lower panel) in vector control (sg-vec) and Paf15-KO (sg-Paf15) NIH3T3 cells. b The levels of genomic DNA methylation (mC) in vector control and Paf15-KO NIH3T3 cells determined by HPLC. Data are presented as mean ± SD from three independent biological replicates. N/C, no significant change (relative difference <5%). c The levels of DNA methylation for each indicated CpG site in mLINE1 in vector control and Paf15-KO NIH3T3 cells were determined by amplicon bisulfite sequencing. d WB analysis of Paf15 expression in control (WT) and Paf15-KO monoclonal (Paf15-KO 2#) NIH3T3 cells. LT, long time. e The levels of genomic DNA methylation (mC) in control and Paf15-KO monoclonal NIH3T3 cells were determined by HPLC. LT, long time. Data are presented as mean ± SD from three technical replicates. N/C, no significant change (relative difference <5%). f WB analysis of Paf15 expression (upper panel) and T7 endonuclease I-based assessment of Paf15 gene editing (lower panel) in vector control (sg-vec) and Paf15-KO (sg-Paf15) E14 cells. g The levels of genomic DNA methylation (mC) in wild-type, Uhrf1-KO, vector control, and Paf15-KO E14 cells were determined by HPLC. LT, long time. HPLC data are presented as mean ± SD from three independent biological replicates. **** p < 0.0001, as determined by unpaired two-tailed Student’s t-test. N/C, no significant change (relative difference <5%). h As in (g), experiments were performed using 5mC immunostaining. Scale bar, 20 μm.

These results collectively indicate that PAF15 is dispensable for maintenance of DNA methylation in mammalian cells.

PAF15 ubiquitination is not in competition with histone H3 ubiquitination

Our data indicate that loss of PAF15 has no significant effect on DNA methylation. Because both PAF15 ubiquitination and histone ubiquitination have been implicated in recruitment of DNMT1, one possibility is that PAF15 and histone H3 are in competition for UHRF1, and that loss of PAF15 ubiquitination could be compensated by increased H3 ubiquitination. We tested this possibility by comparing histone H3 ubiquitination in control and PAF15-KO HeLa cells. As shown in Fig. 4a, loss of PAF15 had no effect on histone H3 ubiquitination in cells synchronized at the G1/S boundary and in S phase. As the deubiquitinating enzyme USP7 has been shown to counteract PAF15 and H3 ubiquitination^23,36, we further assessed whether loss of PAF15 could promote H3 ubiquitination in USP7-KO HeLa cells in which H3 ubiquitination is elevated. We thus generated USP7-KO and USP7/PAF15 double knockout (USP7/PAF15-DKO) HeLa cell lines by CRISPR-Cas9 (Supplementary Fig. 5a). The cells were synchronized at G1/S boundary and in S phase together with PAF15-KO cells, and the subsequent WB analysis revealed that PAF15 deletion did not alter H3 ubiquitination levels in the S phase even in USP7-KO cells (Fig. 4a).

Fig. 4. PAF15 knockout has no effect on H3 ubiquitination and DNMT1 localization to late-replicating DNA foci.

a WB analysis of PAF15 and histone H3 ubiquitination levels in various HeLa cells synchronized to the G1/S boundary or S phase. Cells were synchronized at the G1/S boundary by aphidicolin and to S phase by releasing from G1/S for 4 hours. Vec, PAF15-KO HeLa cells stably transfected with pPyCAGIP vector; re-WT, PAF15-KO HeLa cells stably transfected with PAF15 expression plasmid; re-K/R, PAF15-KO HeLa cells stably transfected with PAF15-K15R/K24R mutant expression plasmid. LE, lower exposure. b WB analysis showing H3 ubiquitination along S phase was unaltered in PAF15-KO HeLa cells. Control HeLa and PAF15-KO HeLa cells were synchronized at G1/S by aphidicolin and nocodazole treatment and then released into S phase. LE, lower exposure. WB results in (a, b) were consistent across two independent biological replicates. c Representative single cell images of Dnmt1 foci formation in wild-type and PAF15-KO NIH3T3. d Representative images of DNMT1 foci formation in wild-type and Paf15-KO E14 cells. e Representative single cell images of DNMT1 foci formation in USP7-KO and USP7/PAF15-DKO HeLa cells. Cells were pulse-labeled with 10 μM EdU for 15 min to label nascent DNA at different stages of S phase. Intensity profiles of fluorescence at marked positions are shown below the corresponding images. Scale bar, 20 μm. Immunofluorescence staining experiments shown in (c–e) were performed for three independent biological replicates with similar results. Quantitative fluorescence analysis of numerous individual cells is provided in the Supplementary Fig. 5d, e, and h.

To better assess the effect of PAF15 disruption on H3 ubiquitination, we compared the temporal dynamics of H3 ubiquitination along S phase progression in parental and PAF15-KO HeLa cells. Consistent with data in Figs. 1a, b, PAF15 ubiquitination increased in the early S phase (15 min–3 h) before declining, while H3 ubiquitination steadily increased throughout the S phase (Fig. 4b). However, PAF15 knockout affected neither the level nor the kinetics of H3 ubiquitination (Fig. 4b). Furthermore, loss of PAF15 also had no effect on H3 ubiquitination level and dynamics along the S phase in USP7-KO HeLa cells (Supplementary Fig. 5b). Taken together, we conclude that, although PAF15 and histone H3 both are ubiquitination substrates of UHRF1, they are not in competition and thus their ubiquitination events are most likely independent.

PAF15 is not required for DNMT1 focal localization to late-replicating DNA

We next investigated whether loss of PAF15 affects DNMT1 recruitment to replicating DNA. In the early S phase, DNA replication primarily occurs in euchromatin. As a consequence, nascent DNA labeled by a 15 min EdU pulse in NIH3T3 cells released into S phase for 1 h exhibited a diffuse EdU distribution pattern (Fig. 4c, lower panel). As the cell cycle progressed to the late S phase to replicate pericentromeric heterochromatin, EdU pulse-labeled nascent DNA was highly enriched in foci characteristic of pericentromeric heterochromatin, which were well-illustrated by dense DAPI foci in NIH3T3 cells (Fig. 4c, lower panel). During this process, DNMT1 exhibited colocalization with EdU pulse-labeled nascent DNA (Fig. 4c, lower panel), as previously documented^46,47. Similarly, DNMT1 also displayed perfect colocalization with replicating DNA-associated PAF15 (Fig. 4c, upper panel), indicating that DNMT1 is recruited to late-replicating DNA during this period of time. Notably, a comparison of PAF15-KO cells with control cells revealed that the loss of PAF15 had no obvious effect on DNMT1 association with nascent late-replicating DNA foci in late S phase (Fig. 4c and Supplementary Fig. 5c, d). This observation was further confirmed by parallel experiments in E14 cells (Fig. 4d and Supplementary Fig. 5e) and HeLa cells (Fig. 4e and Supplementary Fig. 5f–h). For the localization analysis in Fig. 4e, we used USP7-KO HeLa cells because our previous work demonstrated that USP7 depletion markedly enhances DNMT1’s colocalization with EdU-labeled late-replicating pericentromeric foci²³. Parallel assessment in wild-type HeLa cells is provided in Supplementary Fig. 5f, g. Taken together, we conclude that PAF15 is not required for DNMT1 correct localization to late-replicating pericentromeric heterochromatin foci in late S phase. We acknowledge that DNMT1 focal localization is required but not sufficient for DNA maintenance methylation, as a previous study showed that a DNMT1 ΔRFTS mutant localizes normally to late-replicating foci yet fails to maintain methylation⁴⁸.

Loss of UHRF1-mediated PAF15 ubiquitination does not interfere with DNA methylation

UHRF1 catalyzes PAF15 mono-ubiquitination at K15 and K24³⁴, and a ubiquitination-defective PAF15 mutant (K15R/K24R; for simplicity K/R hereafter) was shown to disrupt DNMT1 recruitment and DNA methylation when ectopically overexpressed in mESCs²⁹. Having observed that PAF15 knockout does not alter global DNA methylation in various cell types, we first examined whether ectopic expression of ubiquitination-deficient PAF15 K/R mutant could affect DNA methylation in somatic cells and mESCs. To this end, we reintroduced wild-type or K/R mutant PAF15 into both Paf15-KO NIH3T3 cells and Paf15-KO mESCs via transfection and obtained pooled cells with stable expression of PAF15 (Supplementary Fig. 6a, b). Both exogenously introduced wild-type PAF15 and K/R mutant were expressed to a much higher level than endogenous PAF15 in parental NIH3T3 cells or mESCs (Supplementary Fig. 6a, b). However, subsequent analysis of global DNA methylation by HPLC revealed that ectopic overexpression of either wild-type or mutant PAF15 had no significant effect on global DNA methylation (Supplementary Fig. 6a, b). Furthermore, accurate WGBS analysis revealed only very mild differences in DNA methylation among parental, Paf15-KO, and Paf15-KO re-expressing either wild-type PAF15 or K/R mutant cells (Supplementary Fig. 6c–l).

To rigorously assess the role of PAF15 ubiquitination in DNA methylation, we introduced K15R and K24R mutations into the endogenous Paf15 gene in E14 mESCs  by CRISPR-Cas9 technology (Supplementary Fig. 7a, b). The successful generation of Paf15-K15R/K24R knock-in cell line (Paf15-K/R(KI)) was confirmed by DNA sequencing (Supplementary Fig. 7c) and the loss of Paf15 mono- and di-ubiquitination, but not Paf15 protein (Fig. 5a). Subsequent HPLC analysis showed that Paf15-K/R(KI) cells had a similar global DNA methylation level as the parental E14 cells (Fig. 5b). More accurate WGBS analysis revealed a very modest reduction of DNA methylation in the Paf15-K/R(KI) E14 mESCs  compared to wild-type E14 cells (59.28% vs. 62.79%) (Fig. 5c). Further analyses of bimodal distribution of CpG methylation percentages (Fig. 5d), levels of CpG methylation in genomic elements (Fig. 5e), in HMD and PMD (Fig. 5f), and in a representative genomic region (Fig. 5g) all indicated that Paf15-K/R(KI) mutation did not result in a significant alteration of DNA methylation.

Fig. 5. UHRF1-catalyzed PAF15 ubiquitination is largely dispensable for DNA methylation in E14 cells.

a WB analysis showing the lack of mono- and di-ubiquitinated Paf15 in Paf15-K/R(KI) cells. Note the presence of Paf15 in Paf15-K/R(KI) cells, indicating the K/R knock-in mutation affects only mono- and di-ubiquitination of Paf15. b Global DNA methylation levels in wild-type and Paf15-K/R(KI) E14 cells determined by HPLC analysis. Data are presented as mean ± SD from six technical replicates. N/C, no significant change (relative difference <5%). c WGBS analysis of global CpG methylation levels in wild-type and Paf15-K/R(KI) E14 cells using one independent biological sample. d Bimodal distribution of CpG methylation levels across the genome from the same WGBS dataset in (c). e CpG methylation levels at CpG islands (CpGi), gene bodies, promoters, and transposable-element (TE) regions from the same WGBS dataset in (c). f DNA methylation levels in HMD and PMD regions from the same WGBS dataset in (c). g Representative genome-browser tracks illustrating DNA methylation profiles from the same WGBS dataset in (c). For box plots in (c, e, and f): lines, median; boxes, 25th–75th percentiles; whiskers, 1.5 × interquartile range from the box. P values from one-way ANOVA with Tukey’s post hoc test.

To further validate our findings, we selected three additional independent Paf15-K15R/K24R knock-in E14 clones for analysis. The correct K/R mutations were confirmed by DNA sequencing, and all three clones were shown to lack mono- and di-ubiquitinated PAF15. Subsequent DNA methylation analysis by HPLC showed wild-type levels of global DNA methylation for these clones (Supplementary Fig. 7d, e).

Unlike H3 ubiquitination, PAF15 ubiquitination is not stimulated by hemi-methylated DNA

A key mechanism ensuring the fidelity of DNA maintenance methylation in mammals is the activation of UHRF1 ubiquitin E3 ligase activity toward histones H3/H2B via its binding of hemi-methylated DNA^18–23. This allows UHRF1 to ubiquitinate nucleosome(s) adjacent to hemi-methylated CpG sites, and ubiquitinated nucleosome(s) in turn recruit DNMT1 to fulfill DNA maintenance methylation only in S phase upon DNA replication. Although previous studies^16,23,29,34 and our data here clearly demonstrate that UHRF1 ubiquitinates both PAF15 and histones, it is unknown whether PAF15 ubiquitination by UHRF1 is also activated by hemi-methylated DNA. To address this issue, we first treated HeLa and USP7-KO HeLa cells with 1 μM GSK-3685032, a DNMT1-specific inhibitor, for two days (Fig. 6a). USP7-KO cells were included because loss of USP7 upregulated both PAF15 and H3 ubiquitination, facilitating the analysis of their response to GSK-3685032 treatment. Treatment with 1 μM GSK-3685032 has been shown to markedly increase intracellular hemi-methylated DNA levels, as a consequence of blocking DNA maintenance methylation activity by DNMT1⁴⁹. Furthermore, GSK-3685032 is known to also induce DNMT1 degradation⁵⁰. We confirmed by WB analysis that addition of 1 μM GSK-3685032 effectively downregulated DNMT1 protein level in both control and USP7-KO HeLa cells (Fig. 6a). We also confirmed that GSK-3685032 treatment resulted in a significantly up-regulated histone H3 ubiquitination in both control and USP7-KO HeLa cells, presumably as a consequence of UHRF1 activation by resulting hemi-methylated DNA upon inhibition and degradation of DNMT1 (Fig. 6a). However, examination of PAF15 ubiquitination revealed that GSK-3685032 treatment substantially reduced rather than enhanced PAF15 ubiquitination (Fig. 6a). Similarly, we found that GSK-3685032 treatment strongly upregulated histone H3 ubiquitination but substantially downregulated PAF15 ubiquitination in HCT116, 293 T, NIH3T3, and mESCs (Fig. 6b, c). Thus, GSK-3685032 treatment apparently strongly promotes H3 ubiquitination but inhibits PAF15 ubiquitination.

Fig. 6. H3 ubiquitination, but not PAF15 ubiquitination, is strongly activated by hemi-methylated DNA.

a WB analysis of PAF15 and H3 ubiquitination in wild-type and USP7-KO HeLa cells with or without GSK-3685032 treatment. Cells were treated with 1 μM GSK-3685032 for 24 h, synchronized at the G1/S boundary via 20 h aphidicolin treatment, and released into S phase for 4 h with sustained GSK-3685032 treatment. LE, lower exposure. b As in (a), except experiments were performed in HCT116 and 293 T cells. c WB analysis of PAF15 and H3 ubiquitination in NIH3T3 and E14 cells with or without GSK-3685032 treatment. Cells were treated with DMSO or 1 μM GSK-3685032 for 24 h before harvesting. WB results in a-c were consistent across two independent biological replicates. d Immunostaining showing GSK-3685032 increases the proportion of cells with robust Uhrf1 foci without altering Paf15 localization. Representative images of Uhrf1, Paf15 and nascent EdU-labeled DNA in NIH3T3 cells treated with DMSO or 1 μM GSK-3685032 for 1 h or 24 h. Fluorescence intensity profiles of marked regions are shown below. Scale bar, 20 μm. e Quantification of cells with Uhrf1 foci (upper) and Paf15 foci (lower) corresponding to (d). Data are presented as mean ± SD from four independent biological replicates for each time point. **** p < 0.0001, ** p < 0.01, * p < 0.05, as determined by unpaired two-tailed Student’s t test. n.s. means no significance (p > 0.05). f Representative images of UHRF1 and PAF15 colocalization in HeLa cells treated with DMSO or 1 μM GSK-3685032 for 15 min or 1 h. Fluorescence intensity profiles and ratios of UHRF1 foci-positive cells among PAF15 foci-positive cells are shown. Scale bar, 20 μm. g WB analysis of Paf15 and H3 ubiquitination in wild-type, Uhrf1-KO, Dnmt1-KO and Dnmt-TKO E14 cells. h WB analysis of Paf15 and H3 ubiquitination in vector-control, UHRF1-KO and DNMT1-KO HeLa cells. UHRF1/DNMT1 knockout in HeLa cells was generated via CRISPR-Cas9 using lentiCRISPRv2-sgRNA constructs, followed by 3 days of 1 μg/mL puromycin selection. LE, lower exposure. WB results in (g, h) were consistent across two independent biological replicates.

Previous studies with immunostaining also showed an elevated UHRF1 focal staining pattern upon DNMT1 knockout⁶, as a result of elevated UHRF1 binding of hemi-methylated DNA accumulated in pericentromeric heterochromatin. We treated NIH3T3 cells with or without 1 μM GSK-3685032 for 1 hour or 24 hours and then pulse-labeled the cells with EdU for 15 min. Subsequent immunostaining revealed that GSK-3685032 treatment increased both the percentage of cells with UHRF1 foci and the intensity of UHRF1 focal staining, especially after 24 hours treatment (Fig. 6d, left panel, and Fig. 6e, upper panel). However, PAF15 foci were detected only in EdU-positive S-phase cells, and the pattern of PAF15 foci was not affected by GSK-3685032 treatment (Fig. 6d, right panel, and Fig. 6e, lower panel). Similar results were observed in HeLa cells (Fig. 6f), further indicating that PAF15 ubiquitination correlates with DNA replication but is uncoupled from DNA hemi-methylation.

To further substantiate that hemi-methylated DNA activates histone H3 but not PAF15 ubiquitination, we compared Paf15 ubiquitination in wild-type, Uhrf1-KO, Dnmt1-KO, and Dnmt1/3a/3b-triple knockout (Dnmt-TKO) E14 mESCs (Fig. 6g). Dnmt1-KO in mESCs is known to accumulate hemi-methylated DNA, whereas Dnmt-TKO results in complete loss of DNA methylation. The representative results in Fig. 6g show that H3 ubiquitination is readily detected in Dnmt1-KO and lost in Dnmt-TKO cells, consistent with the respective accumulation and absence of hemi-methylated DNA in these cells. However, Paf15 ubiquitination was not detected in both cells, revealing a surprising essential role of DNMT1 in PAF15 ubiquitination by UHRF1. In support of a role of DNMT1 in PAF15 ubiquitination, we found that knockout of DNMT1 in HeLa cells resulted in a marked reduction of PAF15 ubiquitination in cells synchronized to S phase (Fig. 6h). As controls, DNMT1 knockout elevated H3 ubiquitination levels, and UHRF1 knockout impaired both H3 and PAF15 ubiquitination (Fig. 6g, h).

Altogether, these results clearly indicate that, unlike H3 ubiquitination, PAF15 ubiquitination is not stimulated by hemi-methylated DNA-bound UHRF1. In addition, PAF15 ubiquitination is highly dependent on DNMT1.

DNMT1 promotes PAF15 ubiquitination in a DNA replication-dependent manner

We noticed in multiple cell lines (HCT116, NIH3T3, and mESCs) we have tested that treatment with GSK-3685032 strongly impaired PAF15 ubiquitination levels (Fig. 6a–c). In addition, GSK-3685032 treatment also downregulated PAF15 protein levels to various extents in different cells (Fig. 6a–c). Consistent with the previous report⁵⁰, we observed that GSK-3685032 treatment also induces DNMT1 degradation (Fig. 6a–c). Furthermore, genetic knockout of Dnmt1 in mESCs not only impaired PAF15 ubiquitination but also potently down-regulated Paf15 protein levels (Fig. 6g). These observations raised the possibility that DNMT1 may control PAF15 ubiquitination as well as gene expression and/or protein stability. Notably, while both pharmacological inhibition and genetic ablation of DNMT1 impaired PAF15 ubiquitination, they invariably enhanced histone H3 ubiquitination (Fig. 6a–c, h). This opposing effect also implies that PAF15 and histone H3 ubiquitination are likely mediated by distinct functional pools of UHRF1, with the former being DNMT1-dependent.

To investigate how DNMT1 regulates PAF15 ubiquitination and protein levels, we first examined whether the decrease in PAF15 protein following DNMT1 inhibition occurred at the transcriptional level. Quantitative RT-PCR in NIH3T3 and E14 cells showed that GSK-3685032 treatment did not reduce Paf15 mRNA levels (Supplementary Fig. 8). We next asked whether DNMT1 regulates PAF15 stability via proteasome-dependent degradation. In both Uhrf1-KO and Dnmt1-KO mESC lines, treatment with the proteasome inhibitor MG132 restored PAF15 protein in a dose-dependent manner, suggesting that loss of DNMT1 led to elevated PAF15 degradation via proteasome (Fig. 7a). The much reduced PAF15 protein level in Uhrf1-KO cells suggests that mono-ubiquitination by UHRF1 is likely to protect PAF15 against proteasome degradation. Notably, MG132 treatment rescued PAF15 protein level but not mono- or di-ubiquitination in Dnmt1-KO cells (Fig. 7a), indicating that DNMT1 is also required for UHRF1-mediated ubiquitination of PAF15.

Fig. 7. DNMT1 promotes PAF15 ubiquitination and stability.

a WB analysis of PAF15 protein levels in wild-type, Uhrf1-KO, Dnmt1-KO and Dnmt-TKO E14 cells treated with 0 (DMSO), 5 or 10 μM MG132 for 4.5 h. b WB analysis of PAF15 ubiquitination levels in HeLa cells treated with 1 μg/mL aphidicolin for the indicated durations. c WB analysis of PAF15 ubiquitination levels in HeLa cells treated with the indicated concentrations of aphidicolin for 1 h. d WB analysis of PAF15 ubiquitination levels in HeLa cells subjected to serum starvation and subsequent re-addition, and treated with either aphidicolin or GSK-3685032 for the indicated times. e WB analysis of PAF15 and histone H3 ubiquitination levels in HeLa cells without or with overexpression of FLAG-DNMT1 and FLAG-DNMT1(C1226S) mutant. Cells were unsynchronized. f As in (e), except cells were synchronized at G1/S transition and collected at different periods of S phase as indicated. g WB analysis of PAF15 and histone H3 ubiquitination levels in DLD1 cells without or with overexpression of EGFP-DNMT1. Cells were synchronized to S phase as described above. h WB analysis of PAF15 and histone H3 ubiquitination in control HeLa and HeLa cells with a knock-in FKBPV tag in DNMT1 gene (FKBPV-DNMT1). Cells were synchronized at G1/S boundary by aphidicolin, and released into S phase, and treated with DMSO or 1 μM dTAGV-1 for 2 h. WB results in a–h were consistent across two independent biological replicates. i The working model illustrating why histone ubiquitination but not PAF15 ubiquitination is critical for DNA maintenance methylation. The maintenance of DNA methylation primarily depends on the UHRF1-catalyzed, hemi-methylated DNA-stimulated local histone ubiquitination, rather than UHRF1-catalyzed, DNA replication fork-associated mobile PAF15 ubiquitination. All graphical elements in Fig. 7i were manually created from scratch by the authors using Microsoft PowerPoint, with no third-party materials, pre-existing templates, or AI-generated content utilized.

Having demonstrated that PAF15 ubiquitination is dependent on both DNMT1 and UHRF1 but is uncoupled from DNA hemi-methylation, we investigated the dependency of PAF15 ubiquitination on DNA replication. Inhibition of DNA replication with aphidicolin abolished PAF15 ubiquitination in HeLa cells in a time- and dose-dependent manner (Fig. 7b, c). To further test the dependence of PAF15 ubiquitination on DNA replication, we first arrested HeLa cells in G0/G1 phase via 96 hours serum starvation and then induced cell cycle progression and DNA replication via refeeding with 10% serum for 4 hours and 10 hours, which represented the periods of S phase entry and DNA replication peak, respectively. Subsequent WB analyses showed that PAF15 ubiquitination was very low in G0/G1 arrested cells and markedly increased upon serum refeeding for 4 h. Notably, serum refeeding-induced increase in PAF15 ubiquitination was completely blocked by aphidicolin (Fig. 7d), thus confirming it is a DNA replication-dependent event. Importantly, addition of GSK-3685032 for 4 h did not significantly affect serum refeeding-induced PAF15 ubiquitination, although prolonged 10 h treatment did reduce PAF15 ubiquitination, likely because 10 h GSK-3685032 also substantially down-regulated DNMT1 protein level (Fig. 7d). As DNMT1 protein level was not reduced upon 4 h serum refeeding and GSK-3685032 treatment, yet blocking of DNA replication by aphidicolin abolished PAF15 ubiquitination, we conclude that PAF15 ubiquitination by UHRF1 is entirely dependent on active DNA replication. In these experiments, prolonged GSK-3685032 treatment again enhanced H3 ubiquitination, highlighting a unique role of hemi-methylated DNA in driving UHRF1-catalyzed H3 ubiquitination.

We next asked whether elevating DNMT1 protein levels could enhance PAF15 ubiquitination. Interestingly, we found that ectopic expression of either wild-type DNMT1 or a catalytically inactive mutant (C1226S) increased PAF15 ubiquitination in HeLa cells (Fig. 7e), indicating that the methyltransferase activity is dispensable. Enhancement of PAF15 ubiquitination by ectopic-expressed DNMT1 was replication-dependent, as it occurred only when cells synchronized at the G1/S boundary were released into S phase (Fig. 7f). Similarly, ectopic expression of DNMT1 in colon cancer DLD1 cell line also resulted in a marked increase of PAF15 ubiquitination (Fig. 7g), confirming that DNMT1 promotes PAF15 ubiquitination.

To test a direct role of DNMT1 in PAF15 ubiquitination, we made use of the dTAG system⁵¹ by knocking in an FKBPV-dTAG module into the N-terminus of DNMT1 in HeLa cells. Successful N-terminal tagging of DNMT1 was confirmed by a molecular weight shift (Fig. 7h). Upon G1/S synchronization and S-phase release, induction of DNMT1 degradation with dTAG for 2 h did not affect PAF15 protein levels but substantially reduced PAF15 ubiquitination (Fig. 7h). This result supports a direct role for DNMT1 in promoting DNA replication-dependent, UHRF1-catalyzed PAF15 ubiquitination.

Discussion

PAF15 and its ubiquitination by UHRF1 have a minimal role, if any, in DNA methylation maintenance

Although PAF15 ubiquitination has been proposed as an alternative mode for recruitment of DNMT1 to support DNA maintenance methylation, especially in early-replicating domains, in this study we provide multiple lines of evidence that PAF15 and its associated ubiquitination have a minimal role, if any, in bulk DNA maintenance methylation. First, we show that knockout of PAF15 in HeLa and HEK293T cells has no significant effect on DNA methylation (Fig. 2). As DNA methylation status in control and PAF15-KO cells were examined by multiple approaches including global level of 5mC measurement by HPLC and LC-MS (Fig. 2b, c, e), amplicon-based bisulfite sequencing of repetitive elements LINE1 and HERV-K (Fig. 2f, g), and WGBS (Fig. 2h–l), the minimal impact on DNA methylation upon loss of PAF15 we observed is unlikely due to technical problem in DNA methylation analyses. Second, we found that knockout of PAF15 in NIH3T3 cells and mESCs also has no significant effect on DNA methylation (Fig. 3). Third, unlike the reported results in mESCs, stable ectopic overexpression of either wild-type PAF15 or ubiquitination-defective PAF15 K15R/K24R mutant in NIH3T3 cells and mESCs has no significant effect on global DNA methylation (Supplementary Fig. 6). Fourth, our WGBS results reveal very subtle alterations in DNA methylation in Paf15-K/R(KI) mutant E14 cells (Fig. 5). In this regard, it has been reported that PAF15-KO mice are viable with only mild hematopoietic abnormalities^52,53. These results strongly argue against a critical role of PAF15 in DNA methylation, as genes important for DNA methylation generally give rise to embryonic lethality or pronounced phenotypes when they are deleted in mice^54–59. Given that PAF15-null mice are viable and fertile, we intend to conclude that PAF15 and its associated ubiquitination are likely to play a minor, if any, function in DNA maintenance methylation. Finally, while all major components of the DNA methyltransferase system, including DNMT1, DNMT3, UHRF1, HELLS/LSH, and CDCA7, are evolutionarily co-evolved⁶⁰, PAF15 is exclusive to metazoans (Supplementary Fig. 9), suggesting that it is unlikely to be an essential component of DNA methylation.

Our results contradict Nishiyama et al.’s report that PAF15 is important for maintenance of DNA methylation in mouse embryonic stem cells²⁹. The discrepancy may be due to differences in experimental conditions. The regular serum plus Lif was used for mESC culture in our study, whereas Nishiyama et al. used Lif plus 2i condition. It is well documented that 2i culture conditions maintain ES cells in a naive state that have global DNA hypomethylation^37–39. It was reported that impaired DNA methylation maintenance is the major cause of global DNA hypomethylation in naive ESCs⁴⁰. We suggest that 2i condition may compromise the analysis of PAF15 function in DNA maintenance methylation.

It is noteworthy that, while our study clearly demonstrates that PAF15 is not essential for bulk DNA methylation in mammalian cells we have tested, our study does not exclude the possibility that PAF15 and PAF15 ubiquitination can contribute to DNA maintenance methylation in cells and/or play a critical role in DNA maintenance methylation in some circumstances.

Two distinct pathways are responsible for UHRF1-mediated PAF15 and histone H3 ubiquitination in S phase

While our study argues against a significant role of PAF15 in DNA maintenance methylation, we confirmed that UHRF1 is responsible for PAF15 ubiquitination in S phase. We also confirmed that PAF15 ubiquitination peaks earlier in S phase, whereas histone H3 ubiquitination increases progressively along S phase (Figs. 1, 4, and Supplementary Fig. 5). The difference in PAF15 and H3 ubiquitination kinetics is most likely due to their distinct ways in association with DNA replication forks and UHRF1. As a replication fork-associated protein, PAF15 is likely enriched at replication sites from the beginning of DNA replication and quickly ubiquitinated by the replication fork-enriched UHRF1. Interestingly, a recent study reported that PAF15 is specifically associated with and involved in lagging-strand but not leading-strand DNA synthesis⁶¹. However, as histones are DNA-locked, only the histone at active replication sites can be ubiquitinated by UHRF1, and its ubiquitination level is therefore proportional to the fraction of the genome that has been replicated, and consequently increases along S phase (also balanced by deubiquitination).

A key finding in our study is that while hemi-methylated DNA strongly activates histone H3 ubiquitination by UHRF1, PAF15 ubiquitination by UHRF1 is not and is strictly linked to DNA replication (Figs. 6, 7). Furthermore, we find that PAF15 ubiquitination by UHRF1 is strongly enhanced by DNMT1 (Figs. 6, 7), presumably due to a protein-protein interaction between PAF15 and DNMT1, which helps bring PAF15 to UHRF1. As hemi-methylated DNA-activated UHRF1 ubiquitinates only histone H3 but not PAF15, it is clear that PAF15 ubiquitination is an event completely separated from UHRF1-mediated histone H3 ubiquitination. This is most likely due to the physical separation of PAF15 and histones, as the former is associated with the fast-moving, mobile DNA replication forks, whereas the latter is present in locally deposited nucleosomes. Accordingly, we propose that PAF15 ubiquitination is catalyzed by a pool of UHRF1 that is associated with PCNA, DNMT1 and perhaps the DNA replication machinery in the replication fork (Fig. 7i, upper panel). In this regard, UHRF1 is known to interact with PCNA. On the other hand, H3 ubiquitination is catalyzed by UHRF1, which binds hemi-methylated DNA generated after DNA replication mainly via its unique SRA domain, which induces UHRF1 conformational changes, leading to activation of its E3 ligase activity toward histones^18–22 (Fig. 7i, lower panel).

Why is UHRF1-mediated, hemi-methylated DNA-activated histone ubiquitination crucial for DNA maintenance methylation?

Currently, a key unsolved issue in DNA methylation is to what extent DNA maintenance methylation by UHRF1/DNMT1 is directly coupled to DNA replication and occurs before chromatin reassembly. Although data from previous studies vary substantially on the kinetics of DNA methylation maintenance, it has become clear that the majority of the DNA methylome is maintained after chromatin assembly^45,62–64. If DNA maintenance methylation is dependent on UHRF1-catalyzed histone ubiquitination, one would assume DNA maintenance methylation to occur mainly after chromatin assembly. Thus, the finding that PAF15 is a major UHRF1 substrate and has an important role in DNA methylation provides an appealing alternative mechanism to account for DNA maintenance methylation before chromatin assembly. However, as shown in this study and discussed above, our data strongly argue against a key role of PAF15 in DNA maintenance methylation, although we could not exclude a minor role of PAF15 in DNA methylation (Figs. 2, 3, 5, and Supplementary Fig. 6). Then, why could UHRF1-catalyzed PAF15 ubiquitination not contribute significantly to DNA maintenance methylation?

In principle, DNA maintenance methylation ultimately depends on the accurate recognition of hemi-methylated CpG sites generated upon DNA replication by DNMT1. Considering how ubiquitinated PAF15 and ubiquitinated histones could respectively target DNMT1 to hemi-methylated CpG sites via binding to DNMT1, two completely different scenarios emerge. As PAF15 is associated with the fast-moving DNA replication forks, ubiquitinated PAF15 could enrich DNMT1 at replication forks, but would not deliver DNMT1 to hemi-methylated CpGs. On the other hand, via binding to hemi-methylated CpG sites, UHRF1 can add ubiquitin to adjacent nucleosomal histones; the locally positioned ubiquitinated nucleosomes in turn can recruit and guide DNMT1 to nearby hemi-methylated CpGs to fulfill DNA maintenance methylation. Our study strongly supports UHRF1-catalyzed histone ubiquitination as the key mechanism for DNA maintenance methylation in mammals. The recent finding that PAF15 is selectively loaded to the lagging strand and excluded from the leading strand also argues against a general role of PAF15 in DNA methylation⁶¹.

In line with the local histone-based pathway as the key pathway for DNA maintenance methylation in mammals, it has been shown that histone modifications such as H3K27me3 and H3R2 methylation, mediated by PRC2 and PRMT6, respectively, can inhibit UHRF1-driven H3 ubiquitination and thus compromise the DNMT1 recruitment and DNA methylation fidelity^43,65. In addition, chromatin remodeler HELLS/LSH and its associated CDCA7 are required for DNA maintenance methylation^60,66–69.

While our data argue against the key role of PAF15 in DNA maintenance methylation, PAF15 has been shown to play roles in DNA replication and DNA repair^32,33. We also note that the replication, cell-cycle, and DNA damage phenotypes of PAF15 perturbation appear to be highly cell-type and context-dependent^31,32,61,70. In HeLa cells, we observe only mild replication retardation (Supplementary Fig. 10a) and near-normal cell proliferation upon PAF15 knockout (Supplementary Fig. 10b). We observed that PAF15 knockout moderately enhanced cisplatin sensitivity in HeLa and E14 cells (Supplementary Fig. 10c, d). Finally, neither Paf15 knockout nor ubiquitination site mutation substantially altered cell cycle distribution in E14 cells (Supplementary Fig. 10e). As PAF15 ubiquitination is catalyzed by UHRF1, enhanced by DNMT1, and coupled with DNA replication, we speculate that PAF15 ubiquitination may play a role in coordinating DNA methylation with DNA replication and possibly other associated processes, including DNA repair (Fig. 7i, right panel). Future study is warranted to test this possibility.

Methods

Cell culture

Human cell lines (293 T, HeLa, DLD-1, HCT116) and mouse NIH3T3 cells were cultured in DMEM (Gibco) supplemented with 10% FBS (Gibco), 100 U/mL penicillin, and 100 μg/mL streptomycin (Gibco). The mouse embryonic stem cell line E14 was maintained in DMEM supplemented with 20% FBS, 1000 U/mL leukemia inhibitory factor (Millipore), 2 mM L-glutamine (Gibco), 1× non-essential amino acids (Gibco), 55 μM β-mercaptoethanol (Sigma), and 1 mM sodium pyruvate (Gibco).

Plasmids and primers

DNMT1 cDNA was subcloned into the pPyCAGIP vector (with FLAG or EGFP tags), and the DNMT1-C1226S (D1m) mutant was generated using site-directed mutagenesis. CRISPR-Cas9 plasmids were constructed by cloning guide RNAs (gRNAs) into the lentiCRISPRv2 vector. Guide RNA sequences and genotyping primers are listed in Supplementary Tables 1 and 2. Plasmids used in this study are available upon request from the corresponding author.

Antibodies and reagents

The primary antibodies used in this study are as follows: UHRF1 (Proteintech, 21402-1-AP; 1:1000 for WB and immunostaining); UHRF1 (Santa Cruz, sc-373750, 1:1000 for WB and immunostaining), CyclinE1 (Proteintech, 11554-1-AP; 1:500 for WB); PAF15 (Santa Cruz, sc-390515, 1:500 for WB and 1:200 for immunostaining); PAF15 (CST, #81533, 1:1000 for WB); H3 (Abcam, ab1791, 1:2000 for WB); β-actin (HUABIO, M1210-2; 1:5000 for WB); H3S10P (HUABIO, ET1601-30, 1:2000 for WB); Lamin A/C (Abcam, ab133256, 1:5000 for WB); GAPDH (Abmart, M20006L, 1:5000 for WB); DNMT1 (HUABIO, ET1702-77; 1:1000 for WB and immunostaining), DNMT1 (in-house, 2B4; 1:500 for WB and immunostaining), USP7 (Abcam, ab108931; 1:2000 for WB) and Flag (Sigma, F1804, 1:5000 for WB). H3K18Ub (kind gift from Dr. Falong Lu, 1:1000 for WB).

The following secondary antibodies from Jackson ImmunoResearch were utilized for WB analysis: Alexa Fluor 680 goat anti-rabbit IgG (111-625-144, 1:10,000); Alexa Fluor 790 goat anti-mouse IgG (115-655-146, 1:10,000). For immunostaining, the following secondary antibodies from Jackson ImmunoResearch were employed: Alexa Fluor 594 goat anti-rabbit IgG (111-585-003, 1:500); Alexa Fluor 488 goat anti-mouse IgG (115-545-003, 1:500).

Other reagents used in this study included: EdU (MCE, HY-118411), DAPI (Sigma), GSK-3685032 (MCE, HY-139664), dTAGV-1 (Tocris, 6914), Nocodazole (MCE, HY-13520) and Aphidicolin (MCE, HY-N6733).

Cell cycle synchronization

Cells were synchronized at G1/S transition using a nocodazole/aphidicolin double-block protocol. Briefly, cells were first treated with 50 ng/mL nocodazole for 18 hours. After nocodazole removal, 1 μg/mL aphidicolin was added, followed by an additional 18 hours incubation to achieve G1/S synchronization. Subsequently, cells were released into fresh medium and allowed to progress into S phase for the indicated time periods. For G2/M phase synchronization, cells were treated with 50 ng/mL nocodazole for 18 hours. Synchronization efficiency was validated by EdU staining analyzed via fluorescence microscopy and PI staining coupled with flow cytometry for DNA content analysis.

Gene knockout by CRISPR-Cas9

LentiCRISPRv2 plasmids containing gRNAs were transfected into cells using liposome-based transfection. To establish heterogeneous knockout pools, transfected cells were selected with 4 μg/mL puromycin for 3 days, followed by maintenance in 1 μg/mL puromycin for approximately 2 weeks. Knockout efficiency was evaluated by western blot analysis and T7 endonuclease I assays. For single-cell clone isolation, transfected cells were selected with 1 μg/mL puromycin for 3 days. Individual clones were then expanded in antibiotic-free medium and verified by both WB analysis and Sanger sequencing.

Gene knock-in by CRISPR-Cas9

For construction of Paf15-K15R/K24R knock-in mutant E14 cell lines, gRNA plasmids (pX458) and donor plasmid (pUC57) were co-transfected into E14 cells using liposome transfection reagent (LipoFiter, HanBio) in 6-cm dishes. Transfected cells were selected with 1 μg/mL puromycin for 7 days. Single-cell clones were isolated, expanded, and verified by both WB analysis and Sanger sequencing.

Immunostaining and EdU labeling

Cells on slides were washed with 1× PBS and fixed with 4% paraformaldehyde at 4 °C for 20 min. Cells were then permeabilized with 1% Triton X-100 in 1× PBS at 4 °C for 20 min, blocked with 5% BSA in 1× PBS at 37 °C for 1 hour, and incubated with primary antibodies overnight at 4 °C. After three washes with 1× PBS, cells were incubated with Alexa Fluor 594 goat anti-rabbit IgG and/or Alexa Fluor 488 goat anti-mouse IgG at 37 °C for 1 hour. Nuclei were counterstained with DAPI. Following three additional 1× PBS washes, slides were visualized using an Olympus IX73 microscope. Images were captured at a resolution of 4080 × 3072 pixels (24-bit RGB color depth, sRGB color space) using an Olympus UPlanSApo 40x/0.90 ∞ /0.11-0.23/FN26.5 objective lens, with a calibrated pixel size of 0.16 μm/pixel. All images were acquired in widefield epifluorescence mode as single-plane 2D images without z-stacking. LED illumination and appropriate excitation/emission filter sets were used for each fluorophore. Fluorescence intensity was quantified using ImageJ. Colocalization was analyzed with the JAcoP plugin in ImageJ, and Pearson’s correlation coefficient (PCC) was determined.

The EdU assay was performed using the Cell-Light™ EdU Fluorescent Detection Kit (RiBoBio, #C10310) with modifications. Briefly, cells were labeled with 10 μM EdU (5-ethynyl-2’-deoxyuridine) at 37 °C for 15 min, washed twice with 1× PBS, and fixed with 4% paraformaldehyde at 4 °C for 20 min. After permeabilization with 1% Triton X-100 in 1× PBS at 4 °C for 20 min, EdU was detected using Apollo reaction buffer at 37 °C for 30 min. Subsequent blocking, antibody incubation, and visualization steps were performed as described above.

For 5-methylcytosine (5mC) detection, fixed cells were treated with 3 N HCl at 37 °C for 30 min, followed by neutralization with 10 mM Tris-HCl (pH 8.0). Subsequent steps including blocking, antibody incubation, and visualization were performed as described for immunofluorescence staining.

HPLC and LC-MS

To prepare genomic DNA, cells were resuspended in cell lysis buffer (10 mM Tris-HCl [pH 7.5], 10 mM EDTA, 10 mM NaCl, 0.5% sarcosyl, 0.1 mg/mL RNase (TIANGEN) and incubated at 37 °C overnight. Proteinase K (Merck) was then added to a final concentration of 0.2 mg/mL and incubated at 65 °C for 12 hours. Genomic DNA was extracted using phenol:chloroform, ethanol-precipitated, and dissolved in ddH₂O.

For HPLC nucleoside quantification, 25 μg of denatured genomic DNA was incubated in 133 μL hydrolysis solution (40 mM NaAc pH 5.2, 1 mM ZnSO₄, 1.5 U/mL nuclease P1 (Wako, 145-08221)) at 37 °C overnight. Subsequently, 5 U FastAP (Thermo Fisher Scientific, EF0651) was added and incubated for an additional 8 hours at 37 °C. A 40 μL aliquot of hydrolysate was analyzed using an Agilent 1260 HPLC system with a Supersil AQ-C18 column (5 μm, 4.6 × 250 mm; Elite HPLC). Deoxycytidine and 5-methyl-2’-deoxycytidine were detected at 280 nm using a UV detector.

For UHPLC-MS/MS nucleoside quantification, purified genomic DNA was digested into nucleosides following a previously established protocol⁷¹. Briefly, genomic DNA was incubated in digestion buffer containing 10 mM Tris-HCl (pH 8.0), 1 mM MgCl₂, 1.0 U Super Nuclease (Sino Biological), 0.02 U snake venom phosphodiesterase I (Sangon), and 1.0 U calf intestine alkaline phosphatase (NEB) at 37 °C for 4 h. Residual enzymes were removed using ultrafiltration tubes (Pall) following digestion. UHPLC-MS/MS analysis was performed using an Agilent 1290 Infinity LC system coupled to an Agilent 6410 triple quadrupole mass spectrometer (Santa Clara, CA, USA). Chromatographic separation was conducted on an Agilent Zorbax Eclipse Plus C18 reversed-phase column. The mobile phase consisted of solvent A (0.1% formic acid in water) and solvent B (methanol). Separation was performed with 95% solvent A and 5% solvent B throughout the separation process. Stable isotope-labeled internal standards were applied for accurate calibration and absolute quantification of dC and 5mC.

Regular cell samples were assayed in three independent biological replicates. For monoclonal cell lines, biological replicates were not feasible owing to clonal homogeneity, and at least three independent technical replicates were performed for each sample.

Bisulfite sequencing and WGBS

For amplicon bisulfite sequencing of repetitive elements LINE1 and HERV-K, bisulfite conversion was performed using the EZ DNA Methylation-Gold™ Kit (Zymo Research) according to the manufacturer’s instructions. Bisulfite-converted DNA was amplified by PCR using TaKaRa Ex Taq HS (Takara). PCR products were purified through agarose gel extraction for amplicon sequencing. Primers used for PCR amplification are listed in Supplementary Table 3.

For WGBS library construction and analysis, genomic DNA was fragmented to 200-400 bp using a Covaris S220 system. Single-stranded DNA fragments were subjected to bisulfite treatment, followed by adapter ligation, size selection, and PCR amplification using the Accel-NGS Methyl-Seq DNA Library Kit. Library quality was assessed using an Agilent 5400 system and quantified by quantitative PCR (qPCR). Paired-end sequencing was performed on an Illumina platform.

Raw sequencing data were quality-filtered using TrimGalore (default parameters) to remove low-quality reads and adapters. Filtered reads were aligned to the human (hg38) or mouse (mm10) reference genome using Bismark, with methylation calling performed by Bismark Methylation Extractor. Genomic annotations (common HMD, common PMD, early- and late-replicating regions) were obtained from published sources^45,72.

Cell viability assay

Cell viability was measured using the Cell Counting Kit-8 (CCK-8; Selleck) according to the manufacturer’s instructions. Cells were seeded in 96-well plates at 3000 to 5000 cells per well and cultured overnight. After treatment with inhibitors for the indicated durations, CCK-8 solution was added and incubated for 1 hour at 37 °C. Absorbance at 450 nm was measured, and data are presented as mean ± SD from biological replicates.

Statistics and reproducibility

Statistical analyses were performed using GraphPad Prism 10.1.2 (GraphPad Software). Data are presented as mean ± SD unless otherwise stated. Experiments including HPLC, LC-MS, and RT-qPCR were independently repeated at least three times with comparable results. The two-tailed Student’s t-test was used for pairwise comparisons. P values < 0.05 were considered statistically significant (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001). For HPLC- or LC-MS-based global DNA methylation analysis, differences of less than 5% between datasets were considered nonsignificant changes (N/C).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

J.W. designed and supervised the project. Jiwen Li guided the experiments. Jialun Li performed most of the experiments. Y.H., M.S., S.Z., F.C., Y.Y., and Z.C. participated in experimental design, performed experiments, and analyzed the data. X.S., W.L., and H.W. performed LC-MS analysis. Jialun Li and J.W. prepared the data figures and wrote the manuscript with input from coauthors. M.B. reviewed and edited the draft. J.W., Jialun Li, Y.H., H.W., and M.B. acquired funding. All authors read and approved the final version of the paper.

Peer review

Peer review information

Nature Communications thanks Jikui Song and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This study was supported by grants from National Key Research and Development Program of China (2023YFA1800402 to J.W. and 2023YFA1800401 to Jialun Li), the National Natural Science Foundation of China (32330020 to J.W., 32470622 to Jialun Li, 32200451 to Y.H. and 22234008 to H.W.), and Narodowe Centrum Nauki (National Science Center, 2018/30/Q/NZ2/00669 to M.B.).

Data availability

The WGBS data generated in this study have been deposited in the Genome Sequence Archive (GSA) database under accession numbers HRA010070 and CRA027433. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77759-8.