DNA hypomethylation promotes UHRF1-and SUV39H1/H2-dependent crosstalk between H3K18ub and H3K9me3 to reinforce heterochromatin states

Abstract

Mono-ubiquitination of lysine 18 on histone H3 (H3K18ub), catalyzed by UHRF1, is a DNMT1 docking site that facilitates replication-coupled DNA methylation maintenance. Its functions beyond this are unknown. Here, we genomically map simultaneous increases in UHRF1-dependent H3K18ub and SUV39H1/H2-dependent H3K9me3 following DNMT1 inhibition. Mechanistically, transient accumulation of hemi-methylated DNA at CpG islands facilitates UHRF1 recruitment and E3 ligase activity toward H3K18. Notably, H3K18ub enhances SUV39H1/H2 methyltransferase activity and, in colon cancer cells, nucleates new H3K9me3 domains at CpG island promoters of DNA methylation-silenced tumor suppressor genes (TSGs). Disrupting UHRF1 enzyme activity prevents H3K9me3 accumulation while promoting PRC2-dependent H3K27me3 as a tertiary layer of gene repression in these regions. By contrast, disrupting H3K18ub-dependent SUV39H1/H2 activity enhances the transcriptional activating and antiproliferative effects of DNMT1 inhibition. Collectively, these findings reveal roles for UHRF1 and H3K18ub in regulating a hierarchy of repressive histone methylation signaling and rationalize a combination strategy for epigenetic cancer therapy.

Keywords: DNA methylation; SUV39H1/H2; UHRF1; epigenetic therapy; histone methylation; histone ubiquitination.

Figure 1.. DNA hypomethylation leads to simultaneous increases in H3K18ub and H3K9me3

(A, B) Volcano plots from MS of histone PTMs in RKO cells treated with 300 nM DAC (A) or 1 μM GSK862 (B) for 3 days. Red/blue dots represent significant increases/decreases (p < 0.05) compared to vehicle. Data shown are averages from biological duplicates. (C) Venn diagram of significantly changed and overlapping histone PTMs from panels Figure 1A–B. (D-I) Western blots of the indicated proteins or PTMs querying whole-cell extracts from: (D-E) cell lines treated with 300 nM DAC or 1 μM GSK862 for 2 days; (F) RKO cells following 3 days of doxycycline-inducible shRNA or siRNA DNMT1 KD; (G) DAC-treated HCT116 xenografts; (H) RKO cells ectopically expressing wild-type (WT), catalytic dead (CatX), double ubiquitin-interacting motif mutant (dbUIM), or empty vector control (EV), that underwent endogenous DNMT1 KD for 3 days; and (I) RKO cells collected at indicated time points following a single treatment with 300 nM DAC on Day 0. See also Figure S1.

Figure 2.. UHRF1 E3 ligase activity is responsible for elevated H3K18ub and H3K9me3 levels

(A, B) Western blots of whole-cell extracts from RKO (A) and WI-38 (B) cells with doxycycline-inducible UHRF1 shRNA (Dox-inducible shUHRF1). Treatment paradigm in Figure S2A. (C, D) MS analyses of (C) H3K9 methylation and (D) H3K18ub levels in Dox-inducible shUHRF1 RKO cells. Histone PTM levels are normalized to the total H3 peptide shown on top. Data are mean ± SD from 3 biological replicates. Unpaired two-tailed Student’s t-test was used for statistical analyses to compare drug-treated samples with control for indicated peptides. * p < 0.05, ** p < 0.01, *** p < 0.001, and **** p < 0.0001. Data are 3 biological replicates. (E) UHRF1 domain mutations (top) and quantification of H3K9me3 signals from western blots of Dox-inducible shUHRF1 RKO cells expressing the indicated UHRF1 transgenes (bottom). Treatment paradigm in Figure S2A. Data are mean ± SD from 3 biological replicates. Western blots are shown in Figure S2E–F. See also Figure S2.

Figure 3.. UHRF1-mediated H3K18ub nucleates new H3K9me3 domains following DNMT1 inhibition

(A) siQ-ChIP peaks with ≥ 2-fold signal increase vs. control following the indicated treatments in Dox-inducible shUHRF1 RKO and WI-38 cells. Analyses were based on conserved peaks from biological duplicates. Drug treatment paradigm in Figure S2A. (B) Heatmap of ChromHMM 18-state enrichment analyses using siQ-ChIP peaks from Figure 3A. (C) Average profile of RKO UHRF1, H3K18ub, and H3K9me3 occupancy across TSS and TES. (D) Average profile of H3K18ub and H3K9me3 occupancy across DNMT1 inhibitor-induced UHRF1 siQ-ChIP peaks (n=4,826) from Figure 3A. (E) Genome browser views of siQ-ChIP profiles in Dox-inducible shUHRF1 RKO cells. (F) Scatterplots showing genome-wide correlations of UHRF1, H3K18ub, and H3K9me3 peak signals in DNMT1i-treated RKO cells. (G) Average profile of WI-38 H3K9me3 occupancy across TSS and TES. (H) Genome browser views of siQ-ChIP profiles in Dox-inducible shUHRF1 WI-38 cells. See also Figure S3.

Figure 4.. Hemi-methylated CpG islands facilitate UHRF1 recruitment and activity after DNMT1 inhibition

(A) Average profile of UHRF1, H3K18ub, and H3K9me3 occupancy across DNMT1i-induced hypomethylated DMRs in Dox-inducible shUHRF1 RKO and WI-38 cells. Treatment paradigm in Figure S2A. (B) CpG feature enrichment bias relative to the composition of the DNMT1i-induced hypomethylated DMRs. EPIC array probes covered by RKO and WI-38 DMR cluster 1 and cluster 2 (Figure S4C) were used for analysis. Red/blue indicates a positive/negative enrichment (% of CpGs more/less than expected). (C) Heatmap showing UHRF1, H3K18ub, and H3K9me3 occupancy across all CpG islands of the human genome (n = 27,158) in Dox-inducible shUHRF1 RKO cells. (D) DNA methylation β value distributions from Figure 4C cluster 1 and cluster 2 EPIC CpG probes. (E) (top) RKO cells were repeatedly treated with 1 μM GSK862 at indicated time points, and (bottom) western blots of whole-cell extracts from RKO cells collected at indicated time points. (F) ChIP-qPCR analyses of the relative enrichment of UHRF1 at candidate loci in RKO cells from Figure 4E. Data represent mean ± SEM from technical duplicates. Statistical analysis was performed comparing DNMT1i-treated samples with Vehicle. (G) UHRF1 domain mutations and western blots of whole-cell extracts from the indicated RKO cells. Treatment paradigm in Figure S2A. (H) ChIP-qPCR analyses of the relative enrichment of UHRF1 at candidate loci from the indicated RKO cells. Data represent mean ± SEM from biological duplicates. See also Figure S4.

Figure 5.. UHRF1-mediated H3K18ub enhances SUV39H1/2 methyltransferase activity

(A) Western blots of chromatin extracts from RKO SUV39H2 KO clones (6 and 11) with doxycycline-inducible shRNA targeting SUV39H1. SUV39H1 KD cells were pre-treated with 20 ng/ml doxycycline for 2 days, followed by 20 ng/ml doxycycline treatment with or without 1μM GSK862 (DNMT1i) for an additional 2 days. Control cells were pre-treated with DMSO for 2 days, followed with or without 1μM GSK862 (DNMT1i) for an additional 2 days. (B) ChIP-qPCR analyses of H3K9me3, H3K18ub, and UHRF1 enrichment at candidate loci in the indicated RKO cells. Chr14orf21 serves as the negative control locus. The drug treatments align with Figure 5A. Data represent mean ± SEM from 3 biological replicates. (C) Western blots of chromatin extracts from Dox-inducible shUHRF1 RKO cells. Treatment paradigm in Figure S2A. (D, E) In vitro bioluminescent assay measuring methyltransferase activity using indicated recombinant H3K9 methyltransferases and nucleosome substrates. Data present mean ± SEM from technical duplicates. (F) Immunoprecipitations for H3K18ub and H3K9me3 from DNMT1 inhibitor-treated RKO cells followed by western blots. Cells were treated with 1 μM GSK862 for 2 days. Red triangle points to 25kDa H3K9me3 and H3K18ub bands; black triangle points to the IgG light chain. (G) Schematic of SUV39H1 domain architecture (top) and AlphaFold predicted model for the interaction between SUV39H1 and ubiquitin (bottom). (H) In vitro bioluminescent assay measuring methyltransferase activity using recombinant his-MBP-SUV39H1 and indicated nucleosome substrates at indicated time points. Data present mean ± SEM from technical duplicates. (I, J) ChIP-qPCR analyses of the relative enrichment of H3K9me3 (I) and HA-3xFlag-tagged SUV39H1 (J) at candidate loci in indicated RKO cells. The drug treatments align with Figure 5A. Data present mean ± SEM from 3 biological replicates (I) and technical duplicates (J). (K) Average profile of H3K9me3 occupancy across DNMT1 inhibitor-induced H3K18ub siQ-ChIP peaks (Figure 3A) in Dox-inducible shUHRF1 RKO cells. Treatment paradigm in Figure S2A. (L, M) ChIP-qPCR analyses of the relative enrichment of H3K9me3 at candidate loci in RKO cells. Data present mean ± SEM from technical duplicates. (L) Drug treatments align with Figure 4E. Statistical analysis was performed comparing DNMT1i-treated samples with Vehicle. (M) RKO cells were treated with non-targeting control (NTC) or siRNA targeting HP1 α,β, and γ for 3 days, and with or without 1 μM GSK862 for 2 days. (N) Schematic of the proposed model of H3K18ub-H3K9me3 crosstalk. See also Figure S5.

Figure 6.. UHRF1 E3 ligase activity limits PRC2 activity at hypomethylated CpG islands

(A) Venn diagram illustrating overlapping H3K27me3 siQ-ChIP peaks with ≥ 2-fold signal increase vs. control following the indicated treatments in Dox-inducible shUHRF1 RKO cells. Conserved peaks from biological duplicates were analyzed. Treatment paradigm in Figure S2A. (B) Heatmap of ChromHMM 18-state enrichment analysis using H3K27me3 siQ-ChIP peaks (n=1,582) with ≥ 2-fold signal increase comparing UHRF1 KD+DNMT1i to DNMT1i alone. (C) Average profile of H3K27me3 occupancy across TSS to TES. (D) CpG feature enrichment bias of EPIC array probes (n=5,195) covered by H3K27me3 siQ-ChIP peaks analyzed in Figure 6B. (E) Average profile of H3K27me3 occupancy across the hypomethylated DMRs (n=5,469) mediated by UHRF1 KD and DNMT1i. (F) Overlap of DMRs marked by UHRF1 KD-induced H3K27me3 (Figure S6E Cluster1, 1,845 DMRs) and DNMT1i-induced UHRF1 (Figure S4C Cluster1, 1,636 DMRs). (G) CpG feature enrichment bias of EPIC array probes covered by DMR groups (i), (ii), and (iii) shown in Figure 6F. (H) Heatmap showing UHRF1, H3K9me3, and H3K27me3 occupancy across all CpG islands of the human genome (n = 27,158). (I) Genome browser view of UHRF1 and H3K27me3 siQ-ChIP profiles in Dox-inducible shUHRF1 RKO cells. (J) ChIP-qPCR analyses of the relative enrichment of H3K27me3 at candidate loci from the indicated RKO cells. Data represent mean ± SEM from biological duplicates. Statistical analysis was performed comparing UHRF1 mutants with WT UHRF1 in CTL or DNMT1i-treated groups, respectively. See also Figure S6.

Figure 7.. Disrupting H3K18ub-directed SUV39H1/H2 activity enhances DNMT1 inhibitor efficacy in colon cancer cells

(A) Volcano plot showing differentially expressed genes between drug-treated and control RKO SUV39H2 KO clone11 cells with Dox-inducible shSUV39H1. Red, blue, and grey dots represent significantly upregulated, downregulated, and not significantly changed genes, respectively. Data are from 3 biological replicates. Cells were treated with or without 20ng/mL doxycycline or 1 μM GSK862 for 5 days. (B) Venn diagram of overlapping protein-coding genes upregulated in Figure 7A. (C) Average profile of UHRF1, H3K18ub, and H3K9me3 occupancy across TSS of the 846 genes uniquely upregulated by combined SUV39H1 KD and DNMT1i shown in Figure 7B. (D) Row Z-Score heatmap of genes (n=464) of which TSS were occupied by DNMT1i-induced H3K9me3 as shown in Figure 3A. (E) MA plot showing differential gene expression analysis in RKO SUV39H2 KO clone11 cells with Dox-inducible shSUV39H1. Red dots represent exemplified tumor suppressor genes. (F) GSEA summary of HALLMARK (H) and KEGG (K) gene sets upregulated by combined SUV39H1 KD and DNMT1i vs. CTL or either treatment alone. (G) Row Z-Score heatmap of significantly upregulated WNT signaling and EMT negative regulators by combined SUV39H1 KD and DNMT1i. (H) RT-qPCR analysis of gene expression in SUV39H1/H2 KD/KO RKO cells ectopically expressing WT or 3A mutant SUV39H1. Data represents mean ± SEM from biological duplicates. (I-L) Incucyte cell growth measurements of RKO or HCT116 cells with the indicated treatments. 16 images were captured for each well at each time point, and % confluency results present mean ± SEM of the analysis of 16 images. See also Figure S7.