Polycomb-lamina antagonism partitions heterochromatin at the nuclear periphery

The genome can be divided into two spatially segregated compartments, A and B, which partition active and inactive chromatin states. While constitutive heterochromatin is predominantly located within the B compartment near the nuclear lamina, facultative heterochromatin marked by H3K27me3 spans both compartments. How epigenetic modifications, compartmentalization, and lamina association collectively maintain heterochromatin architecture remains unclear. Here we develop Lamina-Inducible Methylation and Hi-C (LIMe-Hi-C) to jointly measure chromosome conformation, DNA methylation, and lamina positioning. Through LIMe-Hi-C, we identify topologically distinct sub-compartments with high levels of H3K27me3 and differing degrees of lamina association. Inhibition of Polycomb repressive complex 2 (PRC2) reveals that H3K27me3 is essential for sub-compartment segregation. Unexpectedly, PRC2 inhibition promotes lamina association and constitutive heterochromatin spreading into H3K27me3-marked B sub-compartment regions. Consistent with this repositioning, genes originally marked with H3K27me3 in the B compartment, but not the A compartment, remain largely repressed, suggesting that constitutive heterochromatin spreading can compensate for H3K27me3 loss at a transcriptional level. These findings demonstrate that Polycomb sub-compartments and their antagonism with lamina association are fundamental features of genome structure. More broadly, by jointly measuring nuclear position and Hi-C contacts, our study demonstrates how compartmentalization and lamina association represent distinct but interdependent modes of heterochromatin regulation.

Supplementary Figure 1 | LIMe-Hi-C measurements of LADs, CpG methylation, and Hi-C contacts are robust a) Immunoblot depicting levels of Flag-NLS-M.CviPI-Lamin B1 after specified amount of time of induction with doxycycline. GAPDH was utilized as a loading control. The experiment was performed once. Source data are provided as a Source Data file. b) Hi-C contact map comparing LIMe-Hi-C (left) to in situ Hi-C (right) data for replicate 1 at 250 kb resolution for a region on chromosome 1. c) Density heatmap comparing published whole genome bisulfite sequencing (WGBS) CpG methylation fraction (x-axis) to replicate-averaged LIMe-Hi-C CpG methylation fraction (y-axis) across 50 kb bins. Published WGBS dataset is specified in Supplementary Table 1. d) Density heatmap comparing principal component 1 in an individual specified replicate (x-axis) to principal component 1 in a different specified replicate (y-axis) across 50 kb bins for + dox LIMe-Hi-C replicates. e) Density heatmap comparing GpC methylation fraction in an individual specified replicate (xaxis) to GpC methylation fraction in a different specified replicate (y-axis) across 50 kb bins for + dox LIMe-Hi-C replicates. f) Density heatmap comparing replicate-averaged GpC methylation fraction (y-axis) to replicateaveraged principal component 1 (x-axis) across 50 kb bins for the -dox LIMe-Hi-C condition (left). Density heatmap comparing replicate-averaged GpC methylation fraction (y-axis) for the -dox LIMe-Hi-C condition to published K562 DamID data (x-axis) across 50 kb bins (middle). Density heatmap comparing replicate-averaged principal component 1 across 50 kb bins for the + dox (y-axis) and -dox (x-axis) LIMe-Hi-C conditions (right). Published DamID dataset is specified in Supplementary Table 1. g) Hi-C contact map comparing + dox (left) and -dox (right) conditions for replicate 1 at 250 kb resolution. h) Density plot depicting the relative density (y-axis) of region sizes (x-axis) for LIMe and DamID LADs. The density function for each LAD type is independently scaled. i) Observed -expected per-read CpG co-methylation (left) and GpC co-methylation (right) fraction for both arms of chromosome 1 (see Methods). Regions with low coverage near the centromere are marked in white.
Supplementary Figure 2 | Sub-compartments identified in K562 are epigenomically distinct a) Genome browser tracks of lamina GpC methylation, principal component 1 and CpG methylation for LIMe-Hi-C replicates. b) Density plot depicting the relative density (y-axis) of fraction CpG methylation (excluding CpG islands) (x-axis) for 50 kb bins across compartments. The density function for each compartment is independently scaled. c) Boxplot of LIMe-Hi-C features (y-axis: fraction GpC methylation (left), fraction CpG methylation (middle), and principal component 1 (right)) across sub-compartments (x-axis). d) Boxplot showing Polycomb factor enrichment (y-axis, fold-change over input/median signal) for 50 kb bins across sub-compartments (x-axis). Published datasets are specified in Supplementary Table 1. e) Genome browser tracks of H3K27me3 ChIP-seq signal with H3K27me3 peaks and subcompartments below. f) Pie chart depicting fraction base pair overlap of sub-compartments with H3K27me3 peaks. g) Boxplot of z-score-normalized lamina GpC methylation (y-axis) for 50 kb bins within the B compartment by their overlap with H3K27me3 peaks (x-axis). h) Barplot depicting base pair overlap (y-axis) between sub-compartments (x-axis) and LAD subtypes. ciLAD: constitutive inter-LAD; fiLAD: facultative inter-LAD; fLAD: facultative LAD; cLAD: constitutive LAD. i) Boxplot across 50 kb bins genome-wide for PcG-B and Core-B (x-axis) depicting the log2 ratio of the interval's lamina association status (y-axis) if it is interacting with the PcG-B versus Core-B. P values were calculated by a one-sample one-sided t-test to determine if the mean is less than 0. j) Boxplot showing the normalized chromosome-wide average GpC methylation (y-axis) for loci as a function of the sub-compartment status of the loci's interaction partner for PcG-B and Core-B (x-axis) (see Methods). k) Dotplot showing genome-wide average log2(observed/expected contact frequency) (y-axis) for every replicate across sub-compartments (x-axis). O/E denotes observed/expected. In (c-d, g) p values were calculated by a Mann-Whitney-Wilcoxon two-sided test. In (c-d, g, i-j) the interquartile range (IQR) is depicted by the box with the median represented by the center line. Whiskers maximally extend to 1.5 × IQR (with outliers excluded). P values are annotated as follows: ns: not significant; *: 0.01<p≤0.05; **: 0.001<p≤0.01; ***: 0.0001<p≤0.001; ****: p≤0.0001. Exact p values and the number of datapoints (n) compared are provided in the source data file.

Supplementary Figure 3 | Inhibition of EZH2 depletes H3K27me3 and inhibition of DNMT1
depletes DNA methylation globally a) Dose-response curve of concentration inhibitor (x-axis) versus percent growth relative to vehicle (y-axis) for Flag-NLS-M.CviPI-Lamin B1 cells. Data represent mean ± s.e.m across three technical replicates with the experiment performed once. Source data are provided as a Source Data file. b) Immunoblot for Flag-NLS-M.CviPI-Lamin B1 upon EZH2i, DNMT1i, and DMSO treatments +/doxycycline. GAPDH was utilized as a loading control. The experiment was performed once in duplicate. Source data are provided as a Source Data file. c) Scatterplot comparing replicate-averaged H3K27me3 ChIP-seq levels for DMSO (x-axis) and EZH2i (y-axis) treatments for 50 kb bins. d) Genome browser tracks of H3K27me3 ChIP-seq signal for EZH2i and vehicle treatments. e) Boxplot showing log2 fold-change H3K27me3 ChIP-seq signal between EZH2i and vehicle treatments for 50 kb bins (y-axis) across sub-compartments (x-axis). f) Scatterplot comparing replicate-averaged CpG methylation fraction between DMSO (x-axis) and DNMT1i (y-axis) treatments for 50 kb bins. g) Genome browser tracks of the CpG methylation fraction for LIMe-Hi-C samples.

h) Boxplot showing CpG methylation fraction change between DNMT1i and vehicle treatments for 50 kb bins (y-axis) across sub-compartments (x-axis). i) Pearson correlation heatmap of principal component 1 across 50 kb bins for LIMe-Hi-C samples. j) Scatterplot comparing replicate-averaged CpG methylation fraction between DMSO (x-axis)
and EZH2i (y-axis) treatments for 50 kb bins. k) Boxplot showing CpG methylation fraction change between EZH2i and vehicle treatments for 50 kb bins (y-axis) across sub-compartments (x-axis). l) Boxplots showing the normalized chromosome-wide average GpC methylation (y-axis) for loci as a function of the sub-compartment status of the loci's interaction partner for PcG-B and Core-B regions (x-axis) across treatments (see Methods). In (e, h, k-l) the interquartile range (IQR) is depicted by the box with the median represented by the center line. Whiskers maximally extend to 1.5 × IQR (with outliers excluded). P values were calculated by a Mann-Whitney-Wilcoxon two-sided test and are annotated as follows: ns: not significant; *: 0.01<p≤0.05; **: 0.001<p≤0.01; ***: 0.0001<p≤0.001; ****: p≤0.0001. Exact p values and the number of datapoints (n) compared are provided in the source data file. In (c, f, j), the line of best fit is depicted. b) Genome browser tracks of replicate-averaged z-score-normalized lamina GpC methylation levels, principal component 1, and H3K27me3 ChIP-seq signal for LIMe-Hi-C and ChIP-seq data near the HOXA locus. c) Barplot showing for every sub-compartment (x-axis) the observed/simulated Jaccard overlap (y-axis) for regions gaining lamina contact upon EZH2 inhibition relative to a simulated random distribution. d) Aggregate profile plot of H3K27me3 ChIP-seq signal for vehicle treatment (y-axis) across regions gaining lamina contact upon EZH2 inhibition (x-axis) (see Methods). e) Barplot showing for every sub-compartment (x-axis) the observed/simulated Jaccard overlap (y-axis) for regions losing lamina contact upon EZH2 relative to a simulated random distribution. f) Aggregate profile plot of H3K27me3 ChIP-seq signal for vehicle treatment (y-axis) across LIMe LADs (x-axis). g) Aggregate profile plot of log2 fold-change H3K27me3 ChIP-seq signal between EZH2 inhibition and vehicle treatments (y-axis) across B compartments (left, x-axis) and LIMe LADs (right, x-axis). h) Histogram of the frequency (y-axis) of distances (x-axis) for regions gaining lamina contact upon EZH2 inhibition to A/B or LAD borders (top). Histogram of the observed counts (y-axis) of median distances (x-axis) of randomly shuffled regions to A/B or LAD borders (bottom). i) Histogram as in h for regions gaining lamina contact upon DNMT1 inhibition. j) Aggregate profile plots depicting change in z-score-normalized lamina GpC methylation (yaxis) across LIMe LADs (x-axis) upon inhibitor treatments. k) Boxplot showing the difference in z-score-normalized GpC methylation between EZH2i and vehicle treatments (y-axis) for H3K27me3 peaks within 50 kb of compartment borders versus those further than 50 kb away within the A compartment (x-axis). In (a, k) the interquartile range (IQR) is depicted by the box with the median represented by the center line. Whiskers maximally extend to 1.5 × IQR (with outliers excluded). P values were calculated by a Mann-Whitney-Wilcoxon two-sided test and are annotated as follows: ns: not significant; *: 0.01<p≤0.05; **: 0.001<p≤0.01; ***: 0.0001<p≤0.001; ****: p≤0.0001. Exact p values and the number of datapoints (n) compared are provided in the source data file.
Supplementary Figure 5 | LIMe-ID can be used to profile lamina association and reveals that EED inhibition recapitulates the effects of EZH2 inhibition a) Genome browser tracks of GpC and CpG methylation fraction for vehicle treatment for LIMe-ID duplicates with sub-compartment designation below. b) Aggregate profile plot of LIMe-ID replicate-averaged z-score-normalized GpC methylation (yaxis) across B compartments (x-axis). c) Boxplot showing z-score-normalized GpC methylation change between inhibitor and vehicle treatment for 50 kb bins (y-axis) across sub-compartments (x-axis) for the EZH2 inhibitor LIMe-ID data. d) Boxplot showing z-score-normalized GpC methylation change between inhibitor and vehicle treatment for 50 kb bins (y-axis) across sub-compartments (x-axis) for the EED inhibitor LIMe-ID data. e) Boxplots showing the difference in z-score-normalized GpC methylation between inhibitor and vehicle treatment for 50 kb bins (y-axis) segregated into three equally-sized quantiles by log2 fold-change in H3K27me3 (x-axis) for PcG-B domains for both EZH2 inhibitor LIMe-ID data (left) and EED inhibitor LIMe-ID data (right). f) Aggregate profile plots depicting change in z-score-normalized lamina GpC methylation (yaxis) across LIMe LADs (x-axis) upon EZH2 inhibition (left) and EED inhibition (right) (see Methods). g) Aggregate profile plots depicting change in z-score-normalized lamina GpC methylation (yaxis) across EZH2 LIMe-Hi-C lamina differential regions (x-axis) for EZH2 inhibition (left) and EED inhibition (right) (see Methods). h) Scatterplot comparing CpG methylation fraction between DMSO (x-axis) and EZH2 inhibitor treatment (left, y-axis) and EED inhibitor treatment (right, y-axis) for 50 kb bins averaged across replicates. Equation for the line of best fit is depicted in the plot. In (c-e) the interquartile range (IQR) is depicted by the box with the median represented by the center line. Whiskers maximally extend to 1.5 × IQR (with outliers excluded). P values were calculated by a Mann-Whitney-Wilcoxon two-sided test and are annotated as follows: ns: not significant; *: 0.01<p≤0.05; **: 0.001<p≤0.01; ***: 0.0001<p≤0.001; ****: p≤0.0001. Exact p values and the number of datapoints (n) compared are provided in the source data file.