This study complied with all relevant local ethical regulations. Approval for the use of ICF patients’ fibroblasts was obtained after informed consent from all patients or their families (for minors), in agreement with the Helsinki Declaration (CNIL authorization—908256; 14 October 2008)41,66.

Cell culture and drugs

Cells were maintained at 37 °C in a humidified incubator at 5% CO2. Cell lines with DOX-inducible constructs were cultured with 10% tetracycline-free FBS (Biowest). DLD-1 cells were cultured in DMEM–GlutaMAX medium (Gibco) supplemented with 10% FBS. hTERT RPE-1 cells were cultured in DMEM/F-12 GlutaMAX medium (Gibco) containing 10% FBS and 0.123% sodium bicarbonate. HCT116 cells were cultured in McCoy’s 5A modified medium (Gibco) with 10% FBS and 2 mM l-glutamine (Invitrogen). CHM13 hTERT cells (CHM13) (Magee-Women’s Hospital) were cultured in DMEM/F-12 medium containing 10% FBS and supplemented with glutamine (Thermo Fisher Scientific), nonessential amino acids (Thermo Fisher Scientific), 1 mM sodium pyruvate and insulin-transferrin-selenium (Thermo Fisher Scientific). Primary ICF fibroblasts (ICF3)66 were cultured in DMEM–GlutaMAX medium supplemented with 15% FBS. DOX (Sigma-Aldrich) was dissolved in water and used at 100 ng ml−1 unless stated otherwise. IAA (Sigma-Aldrich) was dissolved in water and used at 500 µm. 5Ph-IAA was a gift from M. Kanemaki (National Institute of Genetics, Japan) and now it is commercially available; it was dissolved in DMSO and used at 1 µm. For details on the generation of cell lines, see Supplementary Methods.

Indirect immunofluorescence

For immunofluorescence of cells in interphase, cells were grown to 60–80% confluence in 12 mm 1.5H glass coverslips (Marienfeld Superior). Cells were fixed with 3.7% formaldehyde, 0.5% Triton X-100 in PBS for 10 min (fixation with extraction), washed thrice with 0.1% Triton X-100 in PBS and blocked for 30 min at room temperature or overnight at 4 °C in blocking buffer (0.2 M glycine, 2.5% FBS and 0.1% Triton X-100). Primary antibody incubation was performed for 2 h at room temperature in 2% BSA, 0.1% Triton X-100 in PBS, with mouse anti-CENP-A (1:500; clone 3-19, ENZO, ADI-KAM-CC006-E), rabbit anti-CENP-B (1 μg ml−1 polyclonal; Abcam, ab25734 or 0.539 μg ml−1; clone EPR24047-64; Abcam, ab259855), guinea pig anti-CENP-C (1:1,000; MBL, PD030), rabbit anti-GFP (1 μg ml−1; ChromoTek, PABG1), mouse anti-FLAG (4 μg ml−1; Sigma-Aldrich, F3165), human anticentromere antibody (1:500; Antibodies Incorporated, 15-235-0001), rabbit anti-Geminin (1:500; clone E5Q9S; Cell Signaling Technology, 52508), rabbit monoclonal anti-CDT1 (1:500; clone EPR17891; Abcam, ab202067), rabbit anti-H3K9me3 (1 μg ml−1; Abcam, ab8898). Cells were washed thrice for 3 min with 0.1% Triton X-100 in PBS before incubation with species-specific secondary antibodies for 1 h at room temperature in 2% BSA, 0.1% Triton X-100 in PBS. Alexa Fluor 488, Cy3 and Alexa Fluor 647 conjugated secondary antibodies (Jackson ImmunoResearch Laboratories) were used at 1:500. Cells were washed thrice with 0.1% Triton X-100 in PBS before DAPI staining (1 μg ml−1 in PBS) for 10 min and subsequent mounting with ProLong Gold Antifade Mountant (Invitrogen). For immunofluorescence of mitotic cells, cells were grown in 18-mm glass coverslips (Marienfeld) to ~60% confluence, blocked in mitosis for 3 h with 0.1 μg ml−1 colcemid (Roche) and subjected to a 7-min-long hypotonic shock in a 60% media, 40% water. The semidry coverslips were then centrifuged for 3 min at 320 rcf to spread the chromosomes and then promptly fixed in 3.7% formaldehyde in PBS for 10 min. Immunofluorescence was carried out as described for interphase cells.

Microscopy, live-cell microscopy and image analysis

Fixed imaging was performed on a DeltaVision Core system (Applied Precision) consisting of an Olympus IX71 inverted microscope equipped with a CoolSNAPHQ2 camera (Photometrics) and a 250-W xenon light source. Images ≤4 μm thick were captured in 0.2 μm z-sections at room temperature using a 100× Olympus UPlanSApo oil-immersion objective (numerical aperture 1.4), and then deconvolved and projected (3D maximum intensity) using DeltaVision’s Softworx software. Alternatively, images were acquired on a Thunder Live Cell Imager (Leica) equipped with a Leica K8 CMOS Black-thinned camera, LED lamps and a DFT 51010 dichroic Cube using the LAS X (v1.4.7, Leica) software. Images were captured using a 100× HCX PL APO oil-immersion objective (numerical aperture 1.4) and processed with Instant Computational Clearing. For live-cell imaging, cells were grown on a 35 mm FluoroDish (World Precision Instruments) with 0.17 mm-thick optical-quality glass bottom and fitted with a four-well silicone insert (Ibidi). Time-lapse images were taken every 10 min for 20 h using an inverted Eclipse Ti-E microscope (Nikon) equipped with a CSU-X1 (Yokogawa) spinning disk integrated in Metamorph software, and a four-laser bench (Gataca systems). Furthermore, ~45 μm z-stacks were acquired (z-step size = 3 μm) with a ×60 CFI Plan Apo VC oil-immersion objective (numerical aperture 1.4). The microscope has a motorized Nano z100 piezo stage (Mad City Lab), a stage top incubator (Tokai Hit) and an EMCCD camera (Evolve, Photometrics).

Fiji67 was used to quantify the fluorescence intensity in the deconvolved/Instant Computational Clearing images. For the quantification of fluorescence intensity at the centromeres of cells in interphase, an automatic analysis was performed by detecting the maximum intensity of CENP-C per nucleus, drawing 0.3 μm radius circles (region of interest (ROI)) and measuring the fluorescence of each channel on each ROI (custom Macro script, available upon request). For the quantification of fluorescence intensity in metaphase cells, circles of 0.84 μm radius were automatically drawn (custom Macro script, available upon request) and manually adjusted to encompass the whole centromere of each chromosome. The fluorescence intensity on each ROI was measured for each channel. For the line-scan quantification, the fluorescence intensity on each channel was measured along >3.5 μm long lines manually drawn across the centromeres. The data were manually phased to align the first peak of CENP-C with each centromere.

COBRA

Genomic DNA was extracted using the NucleoSpin Tissue kit (Macherey-Nagel). Between 0.5 and 2 µg of DNA (equalized quantity for parallel comparisons) was bisulfite converted using the EpiTect bisulfite kit (Qiagen) according to the manufacturer’s protocol. The converted DNA was amplified by PCR using locus-specific primers64 (Supplementary Table 1). PCR reactions were performed with 1.5 mM MgCl2, 200 μm dNTPs, 200 μm of each primer and 0.5 U Platinum Taq DNA polymerase (Invitrogen). The PCR products were digested for 3 h with 10 U of the appropriate enzyme at its digestion temperature (Supplementary Table 1). An equal amount of PCR product was incubated alongside the digestions without enzyme (undigested control). The full digests were loaded in 3% agarose gels. Images were acquired using a ChemiDoc (Bio-Rad) and band intensities were quantified using Fiji. The absolute methylation percentage is calculated from the ratio of the intensity of the methylated (digested) bands divided by the sum of all bands on the same lane (unmethylated + methylated). Bisulfite conversion affects C, 5fC and 5caC, all being considered as unmethylated DNA. Furthermore, 5mC and 5hmC are protected from the conversion and therefore considered methylated.

Nanopore sequencing of CenRICHed DNA and mapping-independent methylation analysis

The enrichment of centromeric DNA (CenRICH) was performed as described before65. Nanopore long-read sequencing was performed after CenRICH. The DNA was treated with the Short Read Eliminator kit (cutoff < 25 kb, Circulomics SKUSS-100-101-01) to remove contamination from shorter DNA fragments. Libraries were prepared using the Library Preparation by Sequencing kit (Oxford Nanopore Technology (ONT)), quantified with Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific) and checked by capillary electrophoresis with a TapeStation 4150 system (Agilent). The sequencing was performed on a Spot-ON Flow Cell (R9.4.1) on a MinION Mk1B device. The raw nanopore data were basecalled using the ONT software Guppy version 4.0 with a high accuracy modified-bases model (dna_r9.4.1_450bps_modbases_5mc_hac.cfg), which at the time could only identify 5mC modification. The resulting data were then converted with the fast5mod utility (https://github.com/nanoporetech/fast5mod), to produce as output the sequences of all nanopore reads and their per-base likelihood of methylation, expressed as a value ranging from 0 to 255. Details of the reads processing can be found in Supplementary Methods.

DiMeLo-seq and Fiber-seq

DiMeLo-seq was performed as described in ref. 11 using a mouse anti-CENP-A antibody (ENZO, ADI-KAM-CC006-E) or a rabbit anti-H3K9me3 (Abcam, ab176916), and replacing pA-Hia5 with a nanobody-Hia5 construct described in ref. 68. The CHM13 population was sequenced with R9 chemistry (Supplementary Methods). The CHM13 single clone and RPE-1 hypermethylation experiments were sequenced with R10 chemistry.

Fiber-seq was performed using the same protocol as DiMeLo-seq, with the following alterations: the primary antibody binding, protein A-Hia5 binding and activation steps were skipped. Instead, following nuclear isolation, nuclei were washed once in Tween-wash, then incubated with 1 µM free Hia5 and 800 µM S-adenosyl methionine in 100 µl activation buffer for 30 min at 37 °C. Supernatant was removed, and nuclei were resuspended in 1× PBS.

Ultrahigh molecular weight gDNA was extracted using the NEB Monarch UHMW DNA Extraction Kit for Tissue (NEB, T3010), with a modified protocol. Nuclei from DiMeLo-seq or Fiber-seq reactions were resuspended in 40 µl of 1× PBS. In a separate tube, 1.8 ml NEB Monarch HMW gDNA Tissue Lysis Buffer and 60 µl proteinase K were premixed, then used to gently resuspend nuclei by pipetting with a wide-bore tip. The sample was then incubated at 56 °C for 10 min. Next, 15 µl of RNase A was added, and the sample was returned to the thermomixer to incubate for ten more minutes at 56 °C, this time with agitation at 650–750 rpm. Two glass beads were added to the sample, and 2.5 ml of isopropanol was added. The tube was gently inverted by hand until a clear, jelly-like substance was seen forming around the glass beads. After 1 min incubation at room temperature, the isopropanol was aspirated from the tube, taking care to avoid the DNA. The beads were washed twice with 2 ml Wash Buffer and then were transferred to a bead retainer with a collection tube beneath, quickly spun to remove liquid and quickly transferred into 200 µl extraction elution buffer from the ONT Ultra-Long Sequencing Kit (ONT SQK-ULK001 for R9 and SQK-ULK114 for R10). DNA was eluted overnight. Beads were transferred to a bead retainer with a clean Eppendorf tube beneath and spun at 1,000g for 1 min. The eluted DNA solution was then brought up to a final volume of 750 µl with extraction elution buffer, and quantified by Qubit (Thermo Fisher Scientific, Q3328).

Samples were prepared for sequencing using an ONT Ultra-Long Sequencing Kit (ONT SQK-ULK001 for R9 and SQK-ULK114 for R10) and sequenced using one PromethION flow cell per sample, achieving estimated N50 values of 50–90 kb and throughput of 33–60 Gb per sample for the CHM13 population, estimated N50 values of 124–127 kb and throughput of 36–44 Gb for the CHM13 single clone CENP-A DiMeLo-seq, estimated N50 values of 120–130 kb and throughput of 12 Gb for the CHM13 single clone H3K9me3 DiMeLo-seq, N50 values of 79–86 kb and throughput 37–78 Gb for the CHM13 single clone Fiber-seq. Adaptive sampling was performed on all R10 experiments to deplete noncentromeric reads by providing a bed file with the genome-wide complement of all α-satellite HOR arrays in chm13v2.0 plus a 100 kb buffer to either side.

Processing of DiMeLo-seq and Fiber-seq data

Nanopore (R10) DiMeLo-seq raw data (.pod5 files) were aligned to the T2T-CHM13v2.0 human reference genome and basecalled using dorado (v.0.8.3) and DNA model dna_r10.4.1_e8.2_400bps_sup@v5.0.0 with the following option specified: –modified-bases 6 mA 5mCG_5hmCG. Native barcoding runs had the following additional option specified: –kit-name SQK-NBD114-24. Aligned reads were filtered for primary alignment and sorted using samtools (v1.16.1). Barcoded runs were demultiplexed with dorado demux (v0.8.3).

Fiber-seq

Centromeric ‘core’ reads were defined as those overlapping T2T-CHM13v2.0 CDR coordinates, as specified in ref. 29. Noncore reads (α-sat) were defined as those overlapping active α-satellite HOR, as specified in chm13v2.0_censat_v2.1.bed (T2T Consortium), but excluding CDR coordinates with a 10 kb buffer. Overlapping reads were isolated using samtools view (v1.16.1) and bedtools intersect (v2.30.0). Modification calls were dynamically thresholded using modkit call-mods (v0.1.13) with default parameters. Calls above the empirically determined threshold were clamped to certainty, while those below were discarded. Nucleosomes were called with the ft add-nucleosomes subcommand with default parameters using fibertools (v0.5.3). Chromatin accessibility was assessed by computing the proportion of methylation-sensitive patches over 50 bp using ft extract (fibertools v0.5.3) and custom Python scripts.

CDR hypermethylation

To assess methylation levels at CDRs, samtools view was used to filter for reads aligning to α-sat HOR sequences, as specified in chm13v2.0_censat_v2.1.bed (T2T Consortium). Reads were converted to fasta format with samtools fasta (v1.16.1), re-aligned to the RPE-1 reference genome33 using minimap2 (v2.26-r1175) with parameters specifying ONT sequencing (-x map-ont), then converted back to bam format and filtered for primary alignment with samtools view. MM/ML tags were repaired by broadcasting modification calls from the original BAM file with modkit repair (v0.1.13). Reads were visualized in IGV (v2.18.2 and v2.19.1). To assess off-target methylation, reads were converted to fastq format with samtools fastq (v1.16.1) with the -T MM, ML option specified to preserve modification calls and re-aligned to the RPE-1 reference with minimap2 with the following options specified: -ax map-ont –eqx -y –MD. Re-aligned reads were converted back to BAM format, filtered for primary alignment and sorted with samtools view and samtools sort, respectively. A bed file defining 10,000 random 1,000 bp intervals across the RPE-1 genome was generated with bedtools random (v2.30.0). Reads corresponding to these random intervals were isolated using bedtools intersect (v2.30.0), and the output was re-indexed with samtools index. To generate rolling profile plots, modkit pileup (v0.1.13) was used to count modifications at each reference position. The resulting bedGraph file was sorted using bedtools (v2.30.0), then converted to a bigWig file with bedGraphToBigWig (v2.10). The bigWig file was used to plot rolling mCpG/CpG values with a custom Python script.

5hmC/5mC kinetics analysis by nanopore sequencing

To quantify the kinetics of 5mCpG-to-5hmCpG-to-CpG conversion, we analyzed a time-course dataset from DLD-1 cells generated using native barcoding nanopore sequencing (R10 chemistry). Sequencing and basecalling were performed with the same parameters described above for DiMeLo-seq. The data were processed using modkit pileup with the –cpg setting enabled, and a CpG modification threshold of 0.8 was applied to classify cytosines as methylated (mCpG), hydroxymethylated (hmCpG) or unmethylated. CpG sites were then filtered to retain only those present in reads aligning within CENP-B boxes and α-sat HORs.

Immunofluorescence for expansion microscopy

Cells grown on 12 mm coverslip (1.5H) were permeabilized for 1 min with 0.1 M PHEM buffer (60 mM PIPES, 25 mM HEPES, 2 mM MgCl2, 10 mM EGTA, pH 6.9) with 0.5% Triton X-100, followed by fixation for 10 min with 4% PFA (Electron Microscopy Sciences, 15710) in PBS. Both pre-extraction and fixation solution were prewarmed to 37 °C. After fixation, coverslips were washed thrice with 0.5% Triton in PBS and blocked with 3% BSA, 0.5% Triton in PBS for 30 min at room temperature. Primary antibodies diluted at 1:100 in 3% BSA were added to the coverslips and incubated for 2 h at room temperature. Next, cells were washed thrice with 0.5% Triton in PBS and incubated with DAPI (1 μg ml−1) and secondary antibodies diluted 1:100 in 3% BSA 0.5% Triton in PBS for 2 h at room temperature. Coverslips were washed thrice with PBS and prepared for ExM.

Expansion microscopy

Stained samples were treated with 0.1 mg ml−1 Acryloyl-X (Thermo Fisher Scientific, A20770) in PBS for 2 h, washed thrice with PBS and incubated for 5 min in monomer solution (1× PBS, 2 M NaCl, 2.5% (wt/wt) acrylamide, 0.15% (wt/wt) N,N′-methylenebisacrylamide and 8.625% (wt/wt) sodium acrylate). Coverslips were placed on top of a drop of 90 μl of freshly prepared gelation solution (monomer solution supplemented with 0.2% (wt/wt) TEMED and 0.2% (wt/wt) APS) and incubated for 1 h at 37 °C. Gels were then incubated in digestion solution (8 U ml−1 proteinase K, 1× TAE, 0.5% TX-100, 0.8 M guanidine HCl) for 2 h at 37 °C and washed in PBS containing DAPI (1 μg ml−1). Full expansion was performed in a 10-cm plate by several washings of 30 min with excess volume of Milli-Q water (~4 to 4.5-fold expansion). Expanded samples were immobilized on 25 mm (1.5H) coverslips covered with 0.01% (wt/vol) poly-l-lysin (Sigma-Aldrich, P8920) and imaged on a Nikon Ti-E motorized microscope equipped with a Zyla 4.2Mpx sCMOS camera (Andor) and 100×1.35-NA objective lens (Nikon). Images were acquired as z-stacks at 0.3 μm intervals. Images from Nikon and Zeiss systems were deconvolved with Huygens Professional (v20.10) using up to 40 iterations of the Classic Maximum Likelihood Estimation algorithm with theoretical PSF. The images were only adjusted in brightness and contrast on raw or deconvoluted data using the Fiji software. The number of subunits per single centromere was manually scored in z-stacks using Fiji.

Immunoblot

Whole cell lysates were prepared by resuspending cell pellets in 100 mM Tris–HCl pH 7.6, 4% SDS, 20% glycerol buffer and by sonication. The lysates were BCA-quantified and denatured at 95 °C in 1× Laemmli sample buffer for 5 min. Proteins were separated by SDS–PAGE using TGX gels (Bio-Rad) and transferred onto nitrocellulose membranes (0.45 μm) using the Trans-Blot Turbo transfer system (Bio-Rad). The membrane was blocked with 5% milk in TBS-T (Tris–buffered saline with 0.1% Tween 20) for 2 h at room temperature and incubated overnight at 4 °C with mouse anti-FLAG (0.8 μg ml−1; Sigma-Aldrich), rabbit anti-GAPDH (1:5,000; clone 14C10; Cell Signaling Technology), rabbit anti-HELLS (0.75 μg ml−1; Proteintech), rabbit anti-CENP-B (1 μg ml−1 polyclonal; Abcam, ab25734) or mouse antivinculin (1:2,000; clone hVIN-1, Sigma-Aldrich). Images were acquired on a ChemiDoc imager (Bio-Rad) and analyzed on Image Lab software (v6.1; Bio-Rad).

Statistics and reproducibility

Statistical analysis of all the graphs was performed using GraphPad Prism 10 (GraphPad Software, www.graphpad.com). Details, including the value of cells and/or of centromeres measured, are reported in the figure legends. Full statistical test results can be found in Supplementary Table 5. The choice of the statistical method was made following the guidelines mentioned in ref. 69. Briefly, when possible GraphPad Prism’s normality test was used to determine whether data followed a normal distribution. Depending on the data distribution, unpaired t test or Mann–Whitney test were used to compare two groups, ordinary one-way ANOVA or Kruskal–Wallis test to compare more than two groups, Fisher’s exact test to compare categorical data and one-sample t test to compare samples to a fixed value. No statistical methods were used to predetermine sample sizes. All experiments were performed using several biological replicates. The number of biological replicates and of measurements for each experiment is indicated in the corresponding figure legend. Several steps were taken to ensure the reproducibility of experimental findings and key results were confirmed using complementary experimental approaches. With respect to data randomization, sample assignment to treatment and control groups, as well as data collection, were performed randomly without prior assessment of any cell culture characteristics. Generally, data collection and analysis were not performed blindly to the experimental conditions; however, quantitative image analyses, which are at the heart of all presented data, were independent of experimental conditions. No data were excluded from the analyses.

Reporting summary

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