DNA methylation influences human centromere positioning and function

Abstract

Maintaining the epigenetic identity of centromeres is essential to prevent genome instability. Centromeres are epigenetically defined by the histone H3 variant CENP-A. Prior work in human centromeres has shown that CENP-A is associated with regions of hypomethylated DNA located within large arrays of hypermethylated repeats, but the functional importance of these DNA methylation (DNAme) patterns remains poorly understood. To address this, we developed tools to perturb centromeric DNAme, revealing that it causally influences CENP-A positioning. We show that rapid loss of methylation results in increased binding of centromeric proteins and alterations in centromere architecture, leading to aneuploidy and reduced cell viability. We also demonstrate that gradual centromeric DNA demethylation prompts a process of cellular adaptation. Altogether, we find that DNAme causally influences CENP-A localization and centromere function, offering mechanistic insights into pathological alterations of centromeric DNAme.

Fig. 1. Inducible centromere-targeted demethylation.

a, Scope of the paper. b, DNA (de)methylation cycle. c, Schematic of the ^DBDTET1^AID active demethylation. α-Sat monomers in blue. Black lollipops represent 5mCpGs, white lollipops represent CpGs and CENP-B boxes are in purple. Panel c is created with BioRender.com. d, Left: representative images of CENP-C and FLAG immunofluorescence in DLD-1 FRT ^DBDTET1^AID cells −/+DOX for 48 h. Scale bars = 5 μm; zooms = 4.84 × 4.84 μm. Right: quantification of FLAG fluorescence intensity at centromeres (defined by CENP-C) and in random nuclear positions. Box plot from the 25th to 75th percentiles, central line at median, ‘+’ at mean, min-to-max whiskers. Each dot represents one region of interest (n = 1,409, 1,348 at centromeres, 490, 346 in random positions, −/+DOX, respectively). Kruskal–Wallis test with Dunn’s multiple comparisons: ****P < 0.0001; NS, not significant. e, Average ± standard error of the mean (s.e.m.) FLAG fluorescence intensity at centromeres and random positions in nine different clones (color coded) of DLD-1 cells transduced with LV ^DBDTET1^AID after 48 h of induction with DOX. Two-tailed unpaired t test: ****P < 0.0001. f, Kinetics of methylated (5mC) α-sat DNA as percentage of total DNA by COBRA. DLD-1 FRT cells with DOX-induced expression of active (TET1) or inactive (dTET1) construct. Bars show s.e.m.; n = 2 for dTET1; n = 3 for TET1. Kruskal–Wallis test with Dunn’s multiple comparisons, *P = 0.0325, day 16 to day 0. g, Left: kinetics of 5mC α-sat DNA by COBRA in DLD-1 LV ^DBDTET1^AID −/+DOX. Clones from (e), one biological replicate except for clones 3 and 6 (n = 3 ±s.e.m.). Right: centromeric demethylation rate (5mC_t + 1 − 5mC_t)/((t + 1) − t). h,k, Mapping-independent methylation calling of nanopore reads at α-sats, CENP-B boxes and pericentromeric HSat2 repeats in DLD-1 cells expressing FRT ^DBDTET1^AID (h) or ^AIDDNMT1 ± DNMT3B^−/− [27] treated for 6 days with IAA (k). i, Kinetics of 5mCG, 5hmCG and CG by nanopore sequencing, normalized to day 0. j, Kinetics of 5mC α-sat DNA by COBRA. ^DBD(d)TET1 induction with DOX, degradation of ^AIDDNMT1 with IAA. n = 3 ± s.d. Kruskal–Wallis test at day 4, ****P < 0.0001. LV, lentivirus; a.u., arbitrary units. Source data

Fig. 2. Centromeric hypomethylation increases CENP-A and CENP-B levels.

a, Schematic of experiments shown in (c–g). b, Affinity curves of CENP-B binding to methylated or unmethylated DNA. c,e, Representative immunofluorescence images in DLD-1 LV ^DBDTET1^AID clone 6 at 4 days −/+DOX + 2 days IAA. Scale bars = 5 μm. c, Staining for CENP-A, CENP-B, and CENP-C. Insets: 2.5 × 2.5 μm zoomed views of boxed regions. e, Staining for geminin, CENP-A, and ACA. d, Kinetics of centromeric CENP-A, CENP-B and CENP-C levels in DLD-1 LV ^DBDTET1^AID clones 3, 6 and FRT ^DBDTET1^AID. Mean ± s.e.m., n = 3 (Extended Data Fig. 2a–c), normalized to day 0. Kruskal–Wallis test with Dunn’s multiple comparisons of last-time points versus untreated. **P = 0.0027, ***P = 0.0002. Gray gradient indicates normalized α-sat methylation level at each timepoint. f, Centromeric CENP-A and ACA levels from (e) in geminin negative (G1) and positive (S-G2) cells normalized to −DOX. n = 126, 117 for G1 cells and 124, and 72 for S-G2 cells, for −/+DOX, respectively. Data from one experiment. g, Left: representative images of immunofluorescence in mitotic chromosomes. Scale bars = 1 μm; zooms = 1.5 × 1.5 μm. Right: line scans quantification of CENP-A, CENP-B and CENP-C across mitotic centromeres. Average ± s.e.m. per cell. Two biological replicates, six cells per condition. n = 146 (untreated (NT)) and 170 (DOX) total centromeres. Two-tailed Mann–Whitney test at peak, ****P < 0.0001. h, CENP-A, CENP-B and CENP-C levels in DLD-1 ^AIDDNMT1 and ^AIDDNMT1/DNMT3B^−/− cells before and after 10 days DNMT1 depletion. Three biological replicates, n = 190 (DNMT1/DNMT3B^+/+), 211 (DNMT3B⁻), 176 (DNMT1⁻) and 183 (DNMT1/3B^−/−) total nuclei, *P = 0.0489. i, Schematic of experiments shown in j and k. j, Aggregate plot of 5mC/C across all reads overlapping CDR boundaries in cells expressing ^DBDMQ1_Q147L^AID (purple) or ^DBDdMQ1^AID (green). k, CENP-A and CENP-B levels before and after construct expression for 4 days. Two biological replicates, n = 152 and 91 total nuclei for ^DBDMQ1_Q147L^AID, 119 and 133 total nuclei for ^DBDdMQ1^AID −/+DOX, respectively. Box plots from the 25th to 75th percentiles, central line at median, ‘+’ at mean, min-to-max whiskers. Small dots represent average centromeric fluorescence intensity of a nucleus, color/shape coded by biological replicate. Large dots represent average of all nuclei per biological replicate. Statistics defined by Kruskal–Wallis test with Dunn’s multiple comparisons. ****P < 0.0001 for all pannels. ACA, anticentromere antibody. Source data

Fig. 3. Centromeric hypomethylation favors CENP-A expansion beyond the CDR and increases the centrochromatin accessibility.

a, Schematic depicting CENP-A targeted DiMeLo-seq experiment. Panel a was created with BioRender.com. b, IGV browser region showing individual CENP-A targeted DiMeLo-seq reads overlapping a CDR within the active α-sat HOR array on chr21 in a CHM13 clone without treatment and after ^DBDTET1 expression (DOX). The same reads (gray rectangles) display 5mC (red) and m6A (blue). c,d, Aggregate plots of 5mC as a fraction of total cytosines centered on CDR centers (c) and CDR boundaries, with left CDR boundaries flipped (d). f,g, Aggregate plots of m6A (CENP-A) as a fraction of total adenines across all DiMeLo-seq reads overlapping CDR boundaries in a CHM13 clone untreated (black) and +DOX (red for 5mC, blue for m6A) centered on CDR centers (f) and CDR boundaries, with left CDR boundaries flipped (g). e,h, Distribution of 5mC (e, red) or m6A (CENP-A; h, blue) across CDRs and beyond 100 kb in a CHM13 clone untreated, +DOX and the ratio between them as a fraction of total C/total A, respectively. Each line of the heatmap represents a CDR/centromere/chromosome. CDRs have been scaled to the same length. i, Schematic of experiments shown in j and k. j, Proportion of MSPs >50 bp at CDRs and at active α-sat HORs excluding CDRs. Box plot from the 25th to 75th percentiles, central line at median, ‘+’ at mean, whiskers from 2.5th to 97.5th percentiles. One biological replicate, n = 670 (CDR −DOX), 2,158 (CDR +DOX), 8,777 (α-sat –DOX) and 26,389 (α-sat +DOX) total MSPs. Two-tailed Mann–Whitney test, ****P < 0.0001. k, Aggregate plots of m6A (H3K9me3) as fraction of total A across all DiMeLo-seq reads overlapping CDR boundaries in a CHM13 clone untreated (black) and +DOX (pink). Centered on CDR boundaries, with left CDR boundaries flipped. For all panels DOX = ^DBD-FLAGTET1 expression for 4 days +DOX + 2 days washout. MSPs, methyltransferase-sensitive patches. Source data

Fig. 4. Hypomethylation alters the centromere structure and causes chromosome mis-segregation and aneuploidy.

a, Schematic of the experiments shown in b–k. b, Real-time cell index of naïve DLD-1 cells (NT) and DOX pretreated (4 or 8 days) +DOX during the assay. Triplicates ± s.e.m. Two-tailed one-way ANOVA with Holm–Šídák’s multiple comparisons test, *P = 0.0352 (NT +DOX versus 4 days +DOX), ***P = 0.0006 (NT +DOX versus 8 days +DOX), *P = 0.0207 (4 days +DOX versus 8 days +DOX). c, Representative colony formation assays of DLD-1 LV ^DBDTET1^AID clones (left), and ratio of colonies formed +DOX/NT (right). Three biological replicates ± s.e.m. Two-tailed one-sample t test: **P = 0.0035 (LV cl.3), ***P = 0.0005 (LV cl.6), **P = 0.0029 (FRT TET1), NS (FRT dTET1). d, Percentage of cells with micronuclei at different demethylation levels. Three biological replicates ± s.e.m. Two-sided Fisher’s exact test: *P = 0.0465; 0.0170 (clone 3 +DOX 4, 8 days, respectively); ****P < 0.0001; ***P = 0.0001 (clone 6). e, Correlation between the demethylation ratio at 0–2 days (Fig. 1g) and the change in percentage of cells with micronuclei (day 2 – day 0) for all clones. DLD-1 LV ^DBDTET1^AID clone 6 (f–i). f, Mitosis presenting errors in time-lapse microscopy. Two experimental replicates ± s.d., n = 121 (0 days DOX), 60 (2 days DOX) and 123 (4 days DOX) total mitoses. Two-sided Fisher’s exact test: ****P < 0.0001, **P = 0.0080. g, Representative images of mitotic errors. Fixed samples. Scale bar = 5 μm. h, Frequency of numerical and structural aneuploidy from scGS. Two-sided Fisher’s exact test: *P = 0.0137; ****P < 0.0001. i, Heatmap of the frequency of chromosome-specific aneuploidies from scGS. j, Representative images and line plots from CENP-A expansion microscopy (ExM 4.5-5×, single z-stack) of two sister centromeres untreated or demethylated (+DOX) for 4 days + 2 days IAA. Scale bar = 1 μm. N.I., normalized intensity. k, Number of quantifiable CENP-A sub-domains per single centromere from ExM (n = 3 biological replicates, 300 total centromeres per condition). Two-sided Fisher’s exact test: ****P < 0.0001 (between DOX and centromeres with >2 CENP-A sub-domains). l, Immunoblot of CENP-B. VINCULIN (VNC) used as loading control after stripping. m, Cell growth of DLD-1 LV ^DBDTET1^AID cl.6 and two derived CENP-B^KO sub-clones, untreated or +DOX. Mean confluence from triplicates ± s.e.m. Kruskal–Wallis test with Dunn’s multiple comparisons, ****P < 0.0001 (to WT +DOX). n, Percentage of cells presenting micronuclei. Three biological replicates ± s.e.m. Two-sided Fisher’s exact test: ***P = 0.0002 (WT versus CENP-B KO clones). ANOVA, analysis of variance; WT, wild type. Source data

Fig. 5. Cellular adaptation upon slow, long-term centromere-targeted demethylation and in an ICF4 model.

a, Scheme of the experiments shown in b–j. b, Kinetics of 5mC as percentage of total α-sat DNA by COBRA. Continuous DOX treatment (magenta) and after washout and addition of auxin (DOX _W/O + IAA). Three biological replicates ± s.e.m. Two-tailed Mann–Whitney test at day 30, P = 0.800. c, Centromeric CENP-A and CENP-B levels in DLD-1 FRT ^DBDTET1^AID cells. Three biological replicates, total nuclei analyzed: n = 182 (0 day DOX), 204 (30 days DOX) and 209 (16 days DOX). Kruskal–Wallis test with Dunn’s multiple comparisons *P = 0.0218, **P = 0.0019, ****P < 0.0001. d, Representative colony assays. e, Ratio of colonies (DOX/NT or DOX_W/O/NT). Three biological replicates ± s.e.m. Two-tailed one-sample t test; for TET1 **P = 0.0021 (16 days DOX), **P = 0.0023 (24 days DOX), **P = 0.0039 (32 days DOX), for dTET1 *P = 0.0417 (16 days DOX). f, CRISPR–Cas9 strategy for tagging HELLS with mNG-AID. g, Immunoblot of HELLS (white triangle) and HELLS^mNG-AID (black triangle). HELLS^mNG-AID is degraded with 5Ph-IAA (24 h). LCL from a patient with ICF4(pU; hypomorphic HELLS mutation). GAPDH as loading control. h, Mapping-independent methylation calling of nanopore reads at α-sats, CENP-B boxes and HSat2 repeats. i, 5mC as percentage of total C by COBRA in six loci from HCT116 cells normalized to HELLS^mNG-AID NT. DKO = DNMT1/DMNT3B double knockout (data partially published in ref. ). Two experimental replicates ± s.e.m. j, Centromeric CENP-A and CENP-B levels in HCT116 HELLS^mNG-AID cells. Three biological replicates, n = 205 (day 0) and 215 (day 56) total nuclei. *P = 0.0211. k, Centromeric CENP-A and CENP-B levels in WT and ICF3 (CDCA7^mut) patient-derived fibroblasts. Three biological replicates, n = 75 (WT) and 66 (ICF3) total nuclei. l, Ratio of 5Ph-IAA/DMSO HCT116 HELLS^mNG-AID colonies formed after 10 days. X axis—total days +5Ph-IAA (pretreatment + assay). Three biological replicates ±s.e.m. Two-tailed one-sample t test, no statistical significance: 0.1087 < P < 0.7903. For c, j and k, box plot shows the 25th to 75th percentiles, with central line at median, ‘+’ indicating mean and whiskers extending from the minimum to the maximum values. Small dots represent the average centromeric fluorescence intensity of a nucleus, colored by biological replicate. Large dots represent the average of all nuclei per replicate. Statistics in j and k defined by two-tailed Mann–Whitney test, ****P < 0.0001. mNG-AID, mNeonGreen and AID; LHA, left homology arm; RHA, right homology arm; LCL, lymphoblastoid cell line; GAPDH, glyceraldehyde-3-phosphate dehydrogenase. Source data

Fig. 6. Model of the molecular and cellular consequences of centromeric hypomethylation.

Centromeric hypomethylation causes an increase in CENP-A and CENP-B binding to the centromeres and allows CENP-A to spread outside the CDR boundary. A rapid, short-term demethylation of the centromeres causes chromosome mis-segregation, aneuploidy and reduced viability. A slow, long-term demethylation leads to cellular adaptation to the new epigenetic status. Refer to the ‘Discussion’ for the details. Figure created with BioRender.com.

Extended Data Fig. 1. Validation of the centromere-targeted TET1-induced demethylation.

(a) Schematics of the DNA constructs for targeted centromeric demethylation with the active catalytic domain (CD) of the TET1 enzyme or the inactive dead (dTET1) version. The constructs contain a cytomegalovirus (CMV) promoter, a tetracycline operator (TetO), the DNA-binding domain of CENP-B (CENP-B_DBD), two nuclear localization signals (NLS), a 3× FLAG tag and a micro-AID tag (named AID). A version of the construct with a tandem P2A–T2A (tPT2A) self-cleaving peptide and EGFP was also generated. Panel a is created with BioRender.com. (b) Immunoblot analysis of the constructs’ expression and degradation (anti-FLAG). GAPDH as loading control. (c) Representative agarose gels of COBRA digestions. RE: restriction enzyme. The asterisk demarks the position of the unmethylated (undigested) PCR fragment, M indicates the methylated (digested) fragments. (d) α-sat, CENP-B box and HSat2 methylation level of individual CpGs from DLD-1 FRT ^DBDTET1^AID cells quantified by pyrosequencing. One biological replicate, technical duplicate. (e) Representative images of CENP-C and FLAG immunofluorescence of DLD-1 FRT ^DBDdTET1^AID cells ±100 ng µl⁻¹ doxycycline (DOX) for 48 h. Scale bar: 5 μm. (f) Quantification of methylated DNA (5mC) as percentage of total DNA by COBRA in DLD-1 cells normalized to before treatment. DOX for 16 days + 2 days IAA (red, for Flp-In), or 4 days DOX + 2 days IAA (LV clones). IAA for 4 days to degrade DNMT1 (blue). Three biological replicates ±s.e.m. DNAme was measured at the pericentromeric HSat2 repeats, the hypermethylated CGI promoters of MAEL (Maelstrom Spermatogenic Transposon Silencer), TDRD6 (Tudor Domain Containing 6) genes and GPR119 (G protein-coupled receptor 119). (g) Kinetic quantification of 5mC as percentage of total C at α-satellites by COBRA in DLD-1 FRT ^DBD (d)TET1^AID cells (from Fig. 1f) alongside LV ^DBDTET1^AID cells clones 3 and 6 normalized to day 0 (no induction). Two biological replicates for dTET1, three for all others ±s.e.m. (h) Chromosome-specific heatmap of α-sat 5mCG kinetics by nanopore sequencing. DLD-1 LV ^DBDTET1^AID clone 6 treated with DOX, normalized to day 0. One biological replicate. Source data

Extended Data Fig. 2. Centromeric protein levels upon demethylation throughout the cell cycle.

(a,b) Centromeric CENP-A, CENP-B and CENP-C levels per nucleus in DLD-1 LV ^DBD-TET1^AID clone 3 (a) or clone 6 (b) after induction with DOX over time and 2 days degradation with IAA. Three biological replicates, n = 171, 158, 162 and 155 total nuclei for clone 3, n = 211, 179, 160 and 178 dor clone 6. cl.3 CENP-A day 4 **p = 0.0078, day 8 **p = 0.0027, CENP-B ***p = 0.0006; cl.6 CENP-C day 4 **p = 0.0038, day 8 ***p = 0.0002. (c) Centromeric CENP-A, CENP-B and CENP-C levels per nucleus in DLD-1 FRT ^DBD(d)TET1^AID after DOX for 8 and 16 days followed by IAA for 2 days. Three biological replicates, n = 282, 122 and 127 total nuclei for TET1, n = 135, 143 and 127 for dTET1. (d) CENP-B CUT&RUN-qPCR quantification with centromere 1, 5, 19 primers. Enrichment relative to IgG control without DOX and normalized to HSat2 repeats. Three biological replicates ±s.e.m. (e) Representative images of Cdt1, CENP-A and ACA immunofluorescence in DLD-1 LV ^DBDTET1^AID clone 6 at 4 days ±DOX + 2 days IAA. Scale bar: 5 μm. (f) Centromeric CENP-A and ACA levels from (e) in Cdt1 positive (G1) and negative (S-G2) cells normalized to –DOX. n = 72 and 73 for G1 cells, n = 137 and 138 for S-G2 cells. One experiment. (g) Representative image of CENP-A, CENP-B and CENP-C immunofluorescence in DLD-1 FRT ^DBDTET1^AID mitotic chromosomes before and after demethylation for 16 days (left). Scale bar: 1 μm. Tukey box-and-whiskers plots of mean centromeric CENP-A, CENP-B and CENP-C levels in DLD-1 FRT ^DBD(d)TET1^AID cells (right). Each dot represents a single centromere. Three biological replicates, n = 5,381 and 5,330 centromeres for TET1, n = 4,071 and 4,136 for dTET1. Large colored dots represent the average signal intensity of all centromeres per biological replicate. Unpaired t test: CENP-A *p = 0.0110, CENP-B *p = 0.0316. (a–c,f) Box plot from the 25th to 75th percentiles, central line at median, “+” at mean, min-to-max whiskers. Small dots represent the average centromeric fluorescence intensity in a nucleus and is color coded by biological replicate. Large dots represent the average of all nuclei per biological replicate. Statistics defined by the Kruskal–Wallis test with Dunn’s multiple comparisons. ****p < 0.0001. Source data

Extended Data Fig. 3. CENP-B knock-down does not prevent CENP-A increase.

(a) Schematics of the experiment. DLD-1 LV ^DBDTET1^AID clone 6 cells were demethylated by induction of the construct for 4 days with 100 ng ml⁻¹ DOX, followed by its degradation with 500 μM IAA for 2 days. Control cells were treated with 500 μM IAA for 6 days. A pool of 3 siRNAs against CENP-B or siRNA control were transfected twice, separated by 3 days. (b) Immunoblot analysis of CENP-B protein levels in cells from (a) after the first 3 days of siRNA treatment. GAPDH is used as loading control. c, Integrated centromeric CENP-A and CENP-B signal intensity per nucleus in DLD-1 LV ^DBDTET1^AID clone 6, treated with siRNA control or against CENP-B, ±DOX normalized to −DOX −siCENP-B. Box plot from the 25th to 75th percentiles, central line at median, ‘+’ at mean, min-to-max whiskers. Each small dot represents the average signal intensity of a single nucleus and is color coded by biological replicate. The large dots represent the average signal intensity of all nuclei analyzed per biological replicate. Three biological replicates, n = 512 (siCtrl + IAA), 362 (siCtrl + DOX), 309 (siCENP-B + IAA), 290 (siCENP-B + DOX). Kruskal–Wallis test with Dunn’s multiple comparisons: ****p < 0.0001. d, Aggregate plot of 5mC/C across 10,000 random intervals 1,000 bp long across the genome in cells expressing ^DBDMQ1_Q147L^AID (purple) or ^DBDdMQ1^AID (green). Source data

Extended Data Fig. 4. CENP-A DiMeLo-seq in a CHM13 population matches the observations from CHM13 single clone.

(a) Browser region showing individual CENP-A targeted DiMeLo-seq reads overlapping a CDR within the active α-sat HOR array on chr11 in CHM13 cells (population, ~70% transduction efficiency) without treatment and after 10 days of ^DBDTET1 expression + 2 days washout. For each sample, the same reads (gray rectangles) are copied to two tracks to display mCG (red) and mA (blue) separately. (b,c) Aggregate plots of 5mC/C (b) and m6A/A (c) across all DiMeLo-seq reads overlapping CDR boundaries, with (DOX, red) and without (NT, blue) ^DBDTET1 expression in CHM13 population. Left CDR boundaries are flipped. (d,e) Aggregate plots of m6A/A. (f,g) Aggregate plots of 5mC/C across all DiMeLo-seq reads overlapping the boundaries of 250 random intervals across the genome, with and without (red and blue, respectively) ^DBDTET1 expression for 10 days in CHM13 population (d,f) or for 4 days in the single clone (e,g) followed by 2 days of washout. (h) Representative images of CENP-C, H3K9me3 and ACA immunofluorescence in mitotic chromosomes in DLD-1 LV ^DBDTET1^AID clone 6 cells (left). Cells were demethylated for 4 days with DOX followed by 2 days IAA. Scale bar: 5 μm (right) CENP-C and H3K9me3 line scans across mitotic centromeres. One biological replicate, 6 cells, n = 132 centromeres (NT) and 142 (DOX). Average ±s.e.m. per cell. Source data

Extended Data Fig. 5. Centromeric demethylation in RPE-1 and DLD-1 cells promotes aneuploidy and decreases cell growth.

(a) Cell cycle profile analysis by flow cytometry after 30 min BrdU incorporation in DLD-1 FRT and LV ^DBDTET1^AID cells. Three experimental replicates ± s.d. Brown–Forsythe and Welch ANOVA with Dunnett’s T3 multiple comparisons test, for results, see Supplementary Table 1. (b) Kinetics of methylated (5mC) α-satellite DNA as percentage of total DNA by COBRA in RPE-1 LV ^DBDTET1 clones R1 to R6 with DOX-induced expression of TET1. One biological replicate per clone. (c) Ratio of colonies formed +DOX/−DOX after 7 days of treatment. Three biological replicates ±s.e.m. Two-tailed one-sample t test: **p = 0.0011 (R1); 0.0049 (R2); 0.0017 (R3); 0.0022 (R4) and ***p = 0.0004 (R6). (d) Cell-growth analysis of the indicated RPE-1 clones in absence (dotted lines) or presence (continuous lines) of DOX. Curves represent the mean confluence from triplicates. (e) Percentage of cells with micronuclei over time of induction of the constructs with DOX. DLD-1 FRT expressing active (TET1) or inactive (dTET1) ^DBDTET1^AID. Three biological replicates ±s.e.m. Kruskal–Wallis test with Dunn’s multiple comparisons: *p = 0.0398 (4 days); 0.0145 (8 days). (f) Mitosis duration calculated from timelapse imaging of individual DLD-1 LV ^DBDTET1^AID clone 6 + H2B-mRFP cells untreated (NT) or treated for 2 or 4 days with DOX. (g) Single-cell whole-genome sequencing of DLD-1 LV ^DBDTET1^AID clone 6 cells fully methylated (4 days IAA) or after demethylation with DOX for 2 days followed by 2 days of construct degradation with IAA. Each row represents an individual cell, chromosomes are plotted as columns and centromere position is marked by dotted lines and highlighted by a circle. Colors correspond to a defined copy-number state, legend on the top right. n = 17 untreated cells, n = 69 demethylated cells. h, Number of copy-number alterations (CNA) per cell following hypomethylation at centromeres and chromosome arms detected in g. Two-sided Fisher’s exact test ****p < 0.0001, ***p = 0.0001. Source data

Extended Data Fig. 6. Centromeric CENP-B accumulation is associated with an increase in genome instability and cell lethality.

(a) Canonical and non-canonical centromeres in DLD-1 LV ^DBDTET1^AID cl.6, untreated (NT) or demethylated (+DOX). Three biological replicates, n = 300 centromeres (NT), 301 (+DOX). Two-sided Fisher’s exact test, ****p < 0.0001 (DOX vs. non-canonical centromeres). (b,c) Distribution of CENP-A sub-domains in (b) canonical n = 279 (NT), 217 (+DOX) and in (c) non-canonical centromeres n = 21 (NT), 84 (+DOX). ‘Other’: >5 or unquantifiable sub-domains. Fisher’s exact test between DOX and centromeres with >2 sub-domains ****p < 0.0001 (canonical), *p = 0.0443 (non-canonical). (d) Schematic of CENP-B protein domains. (e,f) Sanger sequences of (e) CENP-B^KO clone 1 and (f) CENP-B^KO clone 2. (g,i) CENP-B (g) and CENP-A (i) centromeric levels in DLD-1 cells untreated (NT) or +DOX to induce overexpression (OE), normalized to NT. Three biological replicates, n = 211 total nuclei (NT), 207 (DOX) for CENP-B^OE, 227 (NT) and 245 (DOX) for CENP-A^OE. Two-tailed Mann–Whitney test, ****p < 0.0001. (h) Methylated (5mC) α-satellite DNA by COBRA in DLD-1 untreated and +DOX to overexpress CENP-A/CENP-B. Three biological replicates ±s.e.m. Two-tailed Mann–Whitney test, p > 0.9999 (CENP-A^OE), p = 0.1000 (CENP-B^OE). (j) Percentage of cells presenting micronuclei. Three biological replicates. Two-tailed Welch’s t test, p = 0.4681 (CENP-A^OE), **p = 0.0033 (CENP-B^OE). (k) Representative images of colony formation assays in CENP-A/CENP-B overexpressing cells. (l) Ratio of colonies formed after 14 days +DOX compared to NT. Two-tailed one-sample t test P = 0.9769 (CENP-A^OE), *p = 0.0239 (CENP-B^OE). (m) Centromeric γH2A.X intensity per nucleus in EdU⁺ DLD-1 cells, untreated or 4 days +DOX to overexpress full-length CENP-B, normalized to untreated. Three biological replicates, n = 468 total nuclei (0 ng DOX), 420 (2.5 ng) and 380 (100 ng). Kruskal–Wallis test with Dunn’s multiple comparisons: **p = 0.0023, ****p < 0.0001. (n) Percentage of metaphase spreads showing reduced DAPI at the centromere in DLD-1 CENP-B^KO cells without rescue or rescued with full-length (FL) CENP-B or with a mutant lacking the C-terminal dimerization domain (ΔC). Three biological replicates ±s.e.m., n = 133 total mitosis (KO), 135 (FL) and 114 (ΔC). Ordinary one-way ANOVA with Tukey’s multiple comparisons, *p = 0.0158 (–/FL), *p = 0.0119 (FL/ΔC). (g,i,m) Box plot from the 25th to 75th percentiles, central line at median, ‘+’ at mean, min-to-max whiskers. Small dots: average fluorescence intensity of a nucleus. Large dots: average of all nuclei per replicate. Source data

Extended Data Fig. 7. Hypomethylation and elevated centromeric protein levels in an RPE-1 clone with short demethylation and long-term recovery.

(a) Schematic of the experiment. RPE-1 ^DBDTET1 (clone R6; Extended Data Fig. 5b–d) + TP53^KO cells were seeded with or without 100 ng ml⁻¹ DOX for 4 days to induce demethylation. At day 4, doxycycline was washed out. At day 5 untreated and DOX-treated single cells were sorted into 96-well plates. Between days 19 and 26 clones were expanded. Approximately one month after sorting, the single clones were screened by COBRA (b) and by immunofluorescence (c). (b) Representative agarose gel of α-sat COBRA digestions. Restriction enzyme (RE): HpyCH4IV. The asterisk demarks the position of the unmethylated (undigested) PCR fragment, M indicates the methylated (digested) fragments. The percentage of α-sat 5mC to total C is indicated below the corresponding lane. (c) Box plot from the 25th to 75th percentiles, central line at median, ‘+’ at mean, min-to-max whiskers. The dots represent the centromeric fluorescence intensity of CENP-A (red) and CENP-B (blue) averaged per nucleus in long-term RPE-1 clones untreated (NT1) or after demethylation (D2) as described in (a). One biological replicate, 137 nuclei analyzed for NT1, 117 nuclei analyzed for D2. Two-tailed Mann–Whitney test ****p < 0.0001. Source data

Extended Data Fig. 8. Validation and characterization of the CRISPR–Cas9 editing of HELLS in HCT116 cells.

(a) Schematic of the 3’ end of HELLS genomic locus (left). The arrows represent the primers used for screening the knock-in of the mNeonGreen-AID tag. The expected product size is indicated. UTR: untranslated region (right) agarose gel electrophoresis of the wild-type (WT) and edited HELLS locus. (b) Sanger sequence of the PCR product from (a) 5’ (top) and 3’ (bottom) ends of the edited HELLS locus in HCT116 HELLS^mNG-AID cells depicting the in-frame knock-in of the mNeonGreen-AID tags. (c) Representative agarose gels of COBRA of HCT116 HELLS^mNG-AID cells, untreated and after 56 days of HELLS degradation with 1 μM of 5Ph-IAA, and HCT116 cells with double knock-out (DKO) of DNMT1/DMNT3B. For α-sat, Alu and LINE1 restriction enzyme (RE): HpyCH4IV, HSat2 RE: BstBI, MAEL and TDRD6 RE: BstUI. The asterisk (*) demarks the position of the unmethylated (undigested) PCR fragment, M indicates the methylated (digested) fragments. (d) Kinetics of α-sat, CENP-B box and HSat2 methylation level of individual CpGs HCT116 HELLS^mNG-AID cells quantified by pyrosequencing. One biological replicate, technical duplicate. (e) Centromeric CENP-A and CENP-B levels in wild-type (WT) and ICF3 (mutations in CDCA7) patient-derived fibroblasts. Three biological replicates, n = 75 (WT) and 66 (ICF3) total nuclei. Box plot from the 25th to 75th percentiles, central line at median, ‘+’ at mean, min-to-max whiskers. Small dots represent the average centromeric fluorescence intensity in a nucleus and are color coded by biological replicate. Large dots represent the average of all nuclei per biological replicate. Two-tailed Mann–Whitney test, ****p < 0.0001. (f) Representative images of colony formation assays in HELLS^mNG-AID cells. Cells were seeded for the 10-day-long colony assays at days 0, 7, 14, 21, 28, 42 and 56 of pre-treatment with 5Ph-IAA (HELLS degradation) either in DMSO after washout of the 5Ph-IAA (top wells) or in continued HELLS degradation with 5Ph-IAA (bottom wells). Total days of 5Ph-IAA treatment are indicated below. Source data