Experimental workflow of cell-cycle phase sorting and Hi-C in mESCs

(A) The Fucci2 reporter (top) labels cell cycle phases: G1/early S (mCherry-hCdt1; red) and S/G2/M (mVenus-hGeminin; green). Asynchronous mESC cultures expressing this reporter (bottom) were fixed, permeabilized, and DNA-stained for FACS sorting. (B) FACS-sorting strategy for the S and G2 phase cell populations based on DNA content. (C) Sequential gating strategy to isolate G1 subpopulations: selection of Geminin-negative cells (Gate A), followed by gating on 2C DNA content (Gate B; entire G1 population), and final fractionation into early, mid, and late G1 based on progressively increasing levels of Cdt1-mCherry fluorescence intensity (approximately 30% of the total population per fraction). (D) Sorted cells from each defined phase were subject to in-situ Hi-C. EG1, early G1; MG1, mid G1; LG1, late G1; ES, early S; MS, mid S; LS, late S.

Cell-cycle-phased Hi-C reveals a stepwise progression of nuclear compartmentalization

(A) Hi-C contact maps (1-Mb resolution) of chromosome 11 for each cell-cycle phase, with the corresponding A/B compartment profiles (Hi-C PC1) shown below each map. Arrows on the Hi-C maps highlight the progressive outward expansion of contact signal from the diagonal, indicating strengthening of long-range interactions. (B) Hi-C saddle plots showing contact enrichment between 1-Mb genomic bins, where both axes are sorted by Hi-C PC1 value (strongest A to strongest B compartment). The schematic below defines the axis ordering. Overall compartment strength (black numerical values) and specific AA, BB, and AB interaction strengths (white numerical values) are quantified. The color scale represents observed/expected (O/E) contact frequencies in 5-percentile increments. Data shown in panels (A) and (B) are from biological replicate 1 (representative of N=2 biological replicates). (C) Hi-C-based compartment strength dynamics across the cell cycle. Each line represents an independent biological replicate. (D) IGV browser view illustrating the correlation between A/B compartment profiles (asynchronous mESCs) and L1/B1 repeat densities (mouse chromosome 2, 200-kb resolution). (E) Workflow of the L1/B1-EdU FISH experiment (see Methods for details). (F) Representative images of MC12 (top) and CBMS1 mESC (bottom) nuclei at the indicated cell-cycle stages, classified based on DAPI and EdU signals. For each nucleus, two merged views are shown: DAPI with EdU, and L1 with B1. Scale bar = 10 µm. (G) Single-cell scatter plots of EdU versus DAPI intensity for MC12 cells (left) and CBMS1 mESCs (right). Colors indicate cell-cycle phase assignments as described in the Methods. (H) L1/B1 segregation index distributions across cell-cycle stages for MC12 cells (left) and mESCs (right). The segregation index was defined as the negative Pearson’s correlation coefficient between L1 and B1 signals, such that higher values indicate stronger spatial segregation (see Fig. S5B and Methods for details). Data from two biological replicates were merged. n= number of nuclei. Pairwise P-values were obtained using a two-tailed unpaired Student’s t-test. EG1, early G1; MG1, mid G1; LG1, late G1; ES, early S; MS, mid S; LS, late S.

Compartment maturation requires S-phase entry but is independent of active DNA synthesis and cohesin-mediated loop extrusion

(A) Experimental design for time-course Hi-C following treatment with INK-128 (INK). (B) Hi-C contact maps (1-Mb resolution) of INK-128-treated cells (24, 72, and 120 h), with corresponding A/B compartment profiles (Hi-C PC1) shown below each map. (C) Hi-C saddle plot analysis of compartment strength for the data in (B), quantifying the overall strength (numerical values in black) and specific AA, BB, and AB interaction frequencies (numerical values in white). The color scale represents O/E contact frequencies in 5-percentile increments. (D) Experimental design for time-course Hi-C following mitotic arrest with Nocodazole (Noc) and release into a Thymidine (Thy) block. (E) Hi-C contact maps and A/B compartment profiles for the experiment in (D). Cell-cycle stages (labeled in pink) were inferred for each population based on Fucci2 reporter fluorescence from FACS analysis. (F) Hi-C saddle-plot analysis of compartment strength for the data in (E). (G) Principal component analysis (PCA) of Hi-C matrices (1-Mb resolution) from asynchronous cell-cycle phases, Nocodazole-Thymidine-blocked cells (THY), and INK-128-treated time-course samples (INK). PCA sample size=50,000. (H) IGV browser view of a representative region on chromosome 2 showing ChIP-seq tracks for RAD21 signal, RAD21 peaks, CTCF signal, CTCF peaks, dual-occupancy RAD21/CTCF peaks (ChIP-seq data from Hansen et al. 201726), and A/B compartments from asynchronous mESCs (this study; 200 kb resolution). (I) Meta-plots showing genome-wide average insulation scores centered at RAD21/CTCF co-occupied sites (1000 randomly sampled peaks) for late G1 (LG1), early S (ES), and G1/S-arrested (THY) cells (at the indicated time points following release from nocodazole and thymidine treatment). Lower insulation scores indicate stronger local insulation and are interpreted as reflecting stronger cohesin-mediated loop extrusion and loop anchoring. Insulation scores were calculated at two resolutions: left, 40- kb resolution using 200- kb sliding windows (±400 kb from peak center); right, 10- kb resolution using 50- kb sliding windows (±200 kb from peak center). Data shown in panels (B), (C), (E), and (F) are from biological replicate 1 (representative of N=2 biological replicates). Data in (I) is from merged biological replicates (N=2).

S-phase A-compartment consolidation revealed by subcompartment analysis

(A) IGV browser tracks of Calder subcompartments (40-kb resolution) for the entire chromosome 11 across all cell-cycle stages, from early G1 (EG1) to G2. (B) Abundance of each Calder subcompartment rank, quantified as the total number of 40-kb genomic bins per cell-cycle stage. (C) Violin plots with overlaid box plots showing the size distribution of Calder subcompartment domains in late G1 (LG1) versus mid-S (MS) phase. Domains are defined as contiguous stretches of genomic bins having the same subcompartment rank. Statistical significance was determined using the Wilcoxon rank-sum test; ns, not significant; *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001. Data in all panels are from merged biological replicates (N=2).

S-phase A-compartment consolidation involves enhanced long-range contacts and structural reorganization

(A) Cis-by-distance Pentad plots31 for all cell-cycle stages from early G1 (EG1) to G2 phase for short (1–10 Mb), short-mid (10–25 Mb), mid/long (25–50 Mb), and long-range (> 50 Mb) interactions. The value at the center of each plot indicates the mean observed/expected (O/E) contact frequency. (B) Observed/Expected Hi-C matrices of two representative regions (chr2: 20–40 Mb and chr8:117–131 Mb) in late G1 (LG1; left panel) and early S phase (ES; middle panel), with corresponding PC1 compartment profiles. The right panel shows a differential heatmap (ES – LG1, early S – late G1). Purple circles highlight weakened interactions between boundaries and the center of a large A domain, or intra-A interactions; black rectangles highlight decreased interactions between neighboring A domains, or inter-A interactions. Panels (A) and (B) present data from merged biological replicates (N=2) analyzed at 200-kb resolution. (C) Schematic model proposing the “A - peninsula” formation, in which internal regions of large A compartments extend away from their boundaries during S-phase. (D) Genomic positions of BAC probes targeting selected A domains on mouse chromosomes 2 and 15. Left boundary probes are shown in green, middle probes in red, and right boundary probes in magenta. (E) Workflow for assigning cell- cycle stages to nuclei following DNA-FISH. Nuclei were first separated into two groups based on median DAPI signal intensity. Nuclei with above-median DAPI intensity were classified as late S/G2 phase. Nuclei with below-median DAPI intensity were further classified as G1 or early/mid S phase based on the number of FISH spots per allele counted manually (see Methods for details). Representative images for each cell- cycle stage are shown; yellow circles indicate individual alleles. Scale bar = 10 µm. (F) Boxplots of pairwise inter-probe distances (µm) by cell- cycle stage. Left panel: chromosome 2; right panel: chromosome 15. For each chromosome, the three probe pairs are (from left to right): green–magenta, green–red, and red–magenta, as indicated above the boxplots. Data from merged biological replicates (N=2); n= nuclei number. Pairwise comparisons between cell cycle stages were performed using the Wilcoxon rank-sum test. P-values were adjusted for multiple comparisons using the Bonferroni method. ns, not significant; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001.

3D genome modeling from Hi-C recapitulates temporal interphase compartment dynamics

(A) 3D genome structures of chromosomes 2 and 17 across the cell cycle, simulated from Hi-C data using the LorDG Modeler in GenomeFlow32 using a conversion factor of 0.6 (10,000 iterations). (B,C) (Left panels) Quantification of outward extension for compartments A and B, showing the mean shortest Euclidean distance from the domain center to its boundaries (calculated as (d1 + d2)/2), normalized by the number of bins per domain. (Right panels) Quantification of boundary movement, showing the shortest Euclidean distance between the boundaries of adjacent domains, normalized by the number of bins per domain. Analyses are shown for small (1–5 Mb) and large (> 5 Mb) domains, comparing late G1 and early S phases. (D) Distribution of the mean bin-to-bin distance within A (Left panels) and B (Right panels) compartment domains across cell-cycle phases. Data are shown for small (1–5 Mb) and large (> 5 Mb) domains. Boxplots in (B–D) show the median and quartiles. Each point in the scatter represents a single domain (n = total domains). Pairwise comparisons between consecutive phases were calculated using the Wilcoxon rank-sum test, and p-values were adjusted using the Benjamini-Hochberg method. Data are from merged biological replicates (N=2), analyzed at 200-kb resolution.

Model of stepwise 3D genome reorganization during the cell cycle

The model proposes four sequential stages: (1) Chromosome unfolding (G1): Gradual formation of long-range interactions from the compact mitotic state. (2) Compartment maturation (G1/S Transition): An abrupt enhancement of compartmentalization upon S-phase entry, independent of DNA synthesis and cohesin-mediated loop extrusion. This stage is characterized by A-compartment consolidation, accompanied by increased long-range A-A interactions and uniform compaction of the B compartment. (3) Compartment stabilization (S-phase): The matured compartment state is maintained throughout the S phase. (4) Chromosome refolding (G2): Global compaction begins in preparation for mitosis.

Validation of Fucci cell-cycle reporters by imaging, FACS, and time-lapse analysis.

(A) Schematic of the EdU labeling protocol in Fucci mESCs for subsequent FACS or imaging analysis. Steps common to both procedures are shown in black. (B) Representative images of Fucci-expressing nuclei (indicated by white arrows) after EdU treatment and DNA labeling across G1, S, G2, and M phases. Scale bar = 10 µm. (C) FACS analysis of EdU-labeled Fucci mESCs, showing DNA content (left), EdU versus DNA content with gated cell-cycle populations (middle), and the same EdU versus DNA content plot with overlaid Fucci signals (right). Geminin-mVenus (green) accumulates at the G1/S transition, while Cdt1-mCherry (red) is enriched in G1. (D) Schematic of the time-lapse imaging and cell-tracking strategy. (E) Temporal dynamics of Fucci reporters (Geminin-mVenus, green; mCdt1-mCherry, red) in single cells after baseline correction and alignment to cell-cycle start (two frames post cell division). Mitotic cells are excluded. (F) Quantification of total cell-cycle length. The mean duration was estimated to be 10.83 ± 2.49 h. (G) Distribution of time spent in G1 (red) versus S/G2/M (green). G1 length was defined as the interval from mitosis to the peak of mCherry fluorescence, with the remaining cell cycle duration assigned to S/G2/M phases. Data in (E–G) are from tracked cells across 2 biological replicates (n=131).

Quality control and analyses of Hi-C data from cell-cycle-phased mESCs.

(A) Principal component analysis (PCA) of replicate Hi-C matrices. The plot shows PCA performed on Hi-C contact matrices (1-Mb resolution) from biological replicates of cell-cycle-phased samples from asynchronous populations, demonstrating replicate concordance. PCA sample size = 50,000. (B) Pearson correlation matrix and hierarchical clustering of Hi-C PC1 compartment profiles (1-Mb resolution) for biological replicates R1 and R2. (C) Contact decay profiles for all cell- cycle phases, plotted from 25 kb to 50 Mb, illustrating progressive changes in cis-interaction frequencies across genomic distances, including a gradual shift from long-range (> 12 Mb) to short-range (< 1 Mb) interactions during the -G1-to-S phase transition. (D) Representative IGV browser tracks of Hi-C PC1 compartment profiles (200-kb resolution) for chromosomes 2, 7, and 18. (E) Quantification of Hi-C PC1 contribution rates for individual chromosomes throughout interphase for replicate 1 (top) and replicate 2 (bottom) at 200-kb resolution. (F) Contact probability, P(s), plotted against genomic distance on a log-log scale (1-Mb resolution). Data shown in panels (C), (D), and (F) are from biological replicate 1 (representative of N=2 biological replicates). EG1, early G1; MG1, mid G1; LG1, late G1; ES, early S; MS, mid S; LS, late S.

Cell-cycle-phased Hi-C data of individual chromosomes

(A) Hi-C contact maps (1-Mb resolution) of representative chromosomes (chr3, 4, 8, 12, and 19) across cell-cycle phases. (B) Contact probability, P(s), versus genomic distance on a log-log scale (1-Mb resolution) for individual chromosomes. Data are from merged biological replicates (N=2).

Re-analysis of single-cell Hi-C data from Nagano et al. 12 (pseudo-bulk analysis)

(A) Contact probability versus genomic distance. Log-log plots of P(s) for merged single-cell Hi-C data (1-Mb resolution, N=2 biological replicates). Consistent with our findings, G1 phase exhibits the longest interaction range compared to mid- and late S/G2 phases. (B) Principal component analysis (PCA) of Hi-C contact matrices (1-Mb resolution; sample size = 50,000) demonstrates high reproducibility between biological replicates. (C) Hi-C contact maps (1-Mb resolution) of representative chromosomes (chr 8, 11, and 15) across cell-cycle phases of replicate 1. (D) Compartment strength quantification. Hi-C saddle plots for replicate 1 (top) and replicate 2 (bottom) at 1-Mb resolution show overall compartment strength (numeric values in black) and specific A-A, B-B, and A-B interaction frequencies (numeric values in white). The color scale represents observed/expected (O/E) contact frequencies in 5-percentile bins. As in our data, A/B compartment strength peaks during S-phase and diminishes in late S/G2.

L1/B1 segregation analysis across the cell cycle in MC12 cells and mESCs

(A) Representative images of MC12 (top) and CBMS1 mESC (bottom) nuclei following L1/B1-EdU FISH. Scale bar= 10 µm. (B) Workflow for generating the dataset used for quantification of L1/B1 segregation across cell-cycle stages (see Methods for details). C, channel; ROI, region of interest.

Validation of G1/G0 and G1/S arrest and comparison of Hi-C compartments in mESCs

(A-C) Reversible G1/G0 arrest of mESCs by INK-128. (A) Schematic of the time-course cell-cycle arrest experiment with INK-128 (1 µM). INK, INK-128. (B) Representative images of cells under each condition (numbered as in (A)), with corresponding cell-count quantification (mean ± SD, N=3). (C) Cell proliferation following long-term (120 h) INK-128 treatment and release into fresh medium (mean ± SD, N=3). SD, standard deviation. (D) FACS analysis of INK-128-treated (G1/G0-arrested) cells. DNA content (top panel), Geminin-mVenus (log scale) versus DNA content (middle panel), and Cdt1-mCherry (log scale) versus DNA content (bottom panel). Purple gates indicate cell populations sorted for Hi-C. (E) Contact probability, P(s), versus genomic distance for INK-128-arrested cells (log-log scale, 1-Mb resolution). Data are from single biological replicates of independent experiments (N=2). (F) FACS analysis of Nocodazole/Thymidine-treated (G1/S-arrested) cells. Panels are arranged as in (D). Cell-cycle stages (labeled in pink) were inferred for each population based on Geminin and Cdt1 fluorescence. (G) Contact probability, P(s), versus genomic distance for G1/S-arrested cells (log-log scale, 1-Mb resolution). Data are from single biological replicates of independent experiments (N=2). (H) t-SNE analysis with k-means clustering (k=2) of compartment saddle plot data (1-Mb resolution) from asynchronous cell-cycle samples (early G1, late G1, early S, mid S) and synchronized samples (INK-128-treated and Nocodazole + Thymidine-treated). Labels “1” and “2” denote biological replicates 1 and 2, respectively.

Insulation score dynamics at RAD21/CTCF sites during normal cell- cycle progression and upon G1/S arrest

(A) Genome browser snapshots of a representative region on chromosome 2 for the indicated cell cycle stages. RAD21/CTCF dual-occupancy sites are shown as a single track at the top (data from Hansen et al. 201726). For each stage, the following tracks are displayed: Hi-C contact map (40- kb resolution), insulation scores at two resolutions (10- kb resolution using a 50- kb sliding windows and 40- kb resolution using 200-kb sliding windows), and A/B compartment track (200- kb resolution). (B) Meta-plots showing genome-wide average insulation scores centered at RAD21/CTCF sites (1000 randomly sampled peaks) for the indicated cell cycle stages. Insulation was calculated at two resolutions: left, 40- kb resolution using 200- kb sliding windows (±400 kb from peak center); right, 10- kb resolution using 50- kb sliding windows (±200 kb from peak center). (C) Boxplots showing genome-wide insulation scores at RAD21/CTCF dual-occupancy sites for the indicated cell- cycle stages. Median values are shown in blue. Pairwise comparisons between consecutive conditions were performed using the Wilcoxon rank-sum tests; ns, not significant; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001. (D) Same as (A) but showing data for G1/S-arrested cells at the indicated time points following release from nocodazole and thymidine treatment. (E) Boxplots as in (C), but comparing insulation scores between late G1, thymidine-arrested cells, and early S-phase cells. Data are from merged biological replicates (N=2).

Calder subcompartment organization of representative chromosomes across interphase

IGV browser tracks of Calder subcompartments (40-kb resolution) for representative chromosomes (chr 2, 6, 13 and 15) across all cell-cycle stages. Data are from merged biological replicates (N=2).

Quantification of A/B compartment PC1 signal “smoothness” by mean-square gradient (MSG)

(A) Example genomic region showing Hi-C PC1 compartment profiles on IGV, illustrating smoothness of the A-compartment signal in S phase compared with G1. (B) (Top) Two schematic examples of compartment-like signals. The signal on the right is an artificially smoothed version of the signal on the left using a sliding-window average. (Middle) The gradient of each curve is plotted, revealing reduced variation in the smoothed curve, while keeping the same vertical limits. (Bottom) The squared gradient for each curve is plotted. The corresponding mean-square gradient (MSG) values are indicated. The smooth curve shows an MSG over tenfold lower, demonstrating the utility of MSG for quantifying smoothness. (C) Ratio of MSG between strong A (top > 1 SD; SD, standard deviation) and strong B (bottom > 1 SD) compartment signals (Hi-C PC1) for each cell-cycle stage. (D) MSG values for compartments defined by different thresholds. From left to right, compartments are defined with increasingly stringent thresholds: the top versus bottom 10%, 5%, 3%, and 2% of Hi-C PC1 values. Data points represent average MSG values. Data in (C, D) are from merged biological replicates (N=2), analyzed at 200-kb resolution.

Pseudo-bulk Hi-C of single cells reveals conserved A-compartment consolidation during S phase across embryonic development

(A) Selection of single-cell populations from Liu et al.29 (HiRES) data for pseudo-bulk Hi-C analysis. The schematic shows the number of cells and developmental stage for each G1 and mid-S population. E7.5, embryonic day 7.5; ExE, extra-embryonic. (B) Hi-C saddle-plot analysis of compartment strength for data generated from the merged single-cell populations in (A) (1-Mb resolution), quantifying overall strength (numerical values in black) and specific AA, BB, and AB interaction frequencies (numerical values in white). The schematic on the right defines the axis ordering. The color scale represents O/E contact frequencies in 5-percentile increments. MS, mid S. (C) Abundance of Calder subcompartment ranks (200-kb resolution), shown as the total number of genomic bins, comparing G1 and mid-S phases across all four developmental stages.

Representative genomic regions showing A-domain reorganization during S-phase

(A) Hi-C compartment strength for different interaction types (inter-A, intra-A, inter-B, intra-B) across interphase, from early G1 (EG1) to G2 phase. Distributions are shown as violin plots with medians indicated by red bars (each dot represents a single chromosome). Statistical significance between consecutive cell-cycle stages was assessed using pairwise Wilcoxon rank-sum tests. (B) Quantification of the Δ median interaction strength between consecutive cell-cycle stages for all interaction types. Inter-A compartment interactions show the largest increase during the late G1-to-early S-phase transition. (C) Observed/Expected Hi-C matrices and corresponding PC1 compartment profiles for six genomic regions (listed below), comparing late G1 (LG1, left panel) and early S phase (ES, middle panel). Purple arrowheads indicate loss of intra-A compartment signal in early S phase relative to late G1, while black arrowheads mark reduced contact frequency between neighboring A compartments. The right panel shows the differential contact heatmap (ES – LG1, early S – late G1). Region coordinates from top to bottom: chr1: 126–144 Mb; chr4: 38–51 Mb; chr5: 110–133 Mb; chr8: 117–131 Mb; chr11: 71–81 Mb; chr17: 22–38 Mb.

Validation of BAC probe positions and biological replicate analysis of inter-probe distances

(A) IGV browser view of BAC probe positions on the linear chromosome (top) and representative metaphase FISH on CBMS1 mESCs after Colcemid treatment (2 hours) (bottom). FISH signals confirm the expected relative localization of probes on chromosomes 2 (left) and 15 (right). Chromosome numbers are indicated on the metaphase spread. CEN, Centromere; TEL, Telomere. Scale bar = 10 µm. (B) Boxplots showing pairwise inter-probe distances (µm) separated by cell- cycle stage and presented for each biological replicate individually (N=2); n, number of nuclei. Top panels: chromosome 2; bottom panels: chromosome 15. For each chromosome, the three probe pairs are (from left to right): green–magenta, green–red, and red–magenta, as indicated above the boxplots. Pairwise comparisons between cell cycle stages (G1 vs early/mid S and early/mid S vs late S/G2) were performed using the Wilcoxon rank-sum test. P-values were adjusted for multiple comparisons using the Bonferroni method. ns, not significant; *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****P ≤ 0.0001.

Correlation between reconstructed 3D structures and Hi-C data across the cell cycle

Absolute Spearman correlation coefficients between experimentally derived Hi-C contact matrices (200-kb bins) and distances from 3D genome structures reconstructed using the LorDG-3D Modeler in GenomeFlow32 using a conversion factor of 0.6 (10,000 iterations). Line plots show correlations for all chromosomes across different cell-cycle phases.

Simulated 3D dynamics of A and B compartment domains across cell-cycle phases

(A) Comparison of mean bin-to-bin distances between A and B compartment domains (> 1 Mb) across cell-cycle phases. Boxplots show the median and interquartile range. Each point represents a single domain. Pairwise comparisons between consecutive phases were performed using the Wilcoxon rank-sum test, and p-values were adjusted using the Benjamini-Hochberg method. (B,C) Mean bin-to-bin distances for A-compartment domains (B) and B-compartment domains (C) separated by chromosome. Data are presented as mean ± SD (points indicate means and error bars indicate SD). Numbers indicate the total number of domains per chromosome. Data in panels (A–C) are from merged biological replicates (N=2) analyzed at 200-kb resolution.