Long-distance association of topological boundaries through nuclear condensates

Abstract

The eukaryotic genome is partitioned into distinct topological domains separated by boundary elements. Emerging data support the concept that several well-established nuclear compartments are ribonucleoprotein condensates assembled through the physical process of phase separation. Here, based on our demonstration that chemical disruption of nuclear condensate assembly weakens the insulation properties of a specific subset (∼20%) of topologically associated domain (TAD) boundaries, we report that the disrupted boundaries are characterized by a high level of transcription and striking spatial clustering. These topological boundary regions tend to be spatially associated, even interchromosomally, segregate with nuclear speckles, and harbor a specific subset of "housekeeping" genes widely expressed in diverse cell types. These observations reveal a previously unappreciated mode of genome organization mediated by conserved boundary elements harboring highly and widely expressed transcription units and associated transcriptional condensates.

Keywords: chromosome architecture; condensate biology; transcription.

Fig. 1.

The aliphatic alcohol 1,6-HD induces a transient reduction in insulation for a subset of TAD boundaries. (A) Schematics of the experimental strategy. (B) Browser example from chromosome 8 showing PC1 principal component–derived compartments (A, positive values; B, negative values) with a region showing small changes to the A/B compartment only after a 5-min treatment. (C) Schematic representation of the insulation score calculation. Insulation score minima were used to define TAD boundaries. (D) Heat map plotting all insulation scores across all boundaries in the specified time points. A total of 949 significantly affected (HDS) boundaries (Top) were identified after a 5-min treatment in replicates (red arrows) vs. 4,568 unaffected boundaries (Bottom). (E) Quantitation of the interaction score shown in D. (F) Venn diagram of significantly affected boundaries at each time point showing most boundaries affected after 5 min rapidly recover over time.

Fig. 2.

HDS boundaries are characterized by high transcription levels and undergo transcriptional pausing after 1,6-HD treatment. (A) Enrichment of transcription factor ChIP-seq datasets on HDS and unaffected boundaries from the Cistrome database. Each dot on the resulting plot represents a ChIP-seq sample with its corresponding GIGGLE score, where higher GIGGLE scores indicate more enrichment. (B) Heat map showing PRO-seq transcription levels on HDS (Top) or unaffected (Bottom) centered on boundaries at each time point. (C) Quantitation of PRO-seq transcription levels as shown in B. (D) Heat maps of PRO-seq transcription at HDS or unaffected boundaries after a 5-min 1,6-hexanediol treatment in replicates, centering on gene transcription start sites (TSSs) for genes within the given regions. (E) Cumulative pausing ratio plots of all genes in control (2,5-HD) (blue) or 1,6-HD–treated (yellow) RUES1 cells. (F) Pausing ratios of significantly more paused (Left, n = 545) or less paused (Right, n = 77) genes in control (2,5-HD) or 1,6-HD 5-min–treated RUES cells. (G) Relative enrichment of 1,6-HD deregulated genes with increased (up) or decreased (down) pausing overlapping HDS vs. unaffected boundaries. Transcript numbers were normalized to the total genomic size of HDS and unaffected regions.

Fig. 3.

Transcription inhibitors recapitulate insulation reduction on HDS boundaries. (A) Schematic of the mode of action of an inhibitor of transcriptional initiation (triptolide) or transcriptional elongation (flavopiridol). (B) Schematic of the treatment strategy using inhibitors for downstream Hi-C. (C) Heat map plotting insulation scores on significantly affected (HDS, Top) vs. unaffected boundaries (Bottom) in flavopiridol- and triptolide-treated cells. (D) Quantitation of interaction scores shown in C. (E) Plots of Hi-C matrices in control, flavopiridol, or triptolide RUES cells treated with inhibitors for 3 h. Arrows highlight increased long-range interactions after treatment over regions with HDS-boundary (HDSb) clusters.

Fig. 4.

HDS boundaries are organized in clusters that exhibit enhanced interchromosomal interactions and increased association with interchromatin granule clusters (speckles). (A) Whole-genome chromosome schematic shows HDS boundaries (red dots) and clusters (black lines). (B) Example of HDS boundary clusters on chromosome 3. HDS boundaries (red lines), HDS clusters (black bars), and all boundaries (blue lines). (C) Hi-C interaction matrix showing all chromosomes. The black box highlights the location of interchromosomal interactions between chromosomes 1 and 3. (D) Example of Hi-C interaction between chromosomes 1 and 3. Increased interactions (red) are enriched over HDS boundaries (green dashes) and HDS-boundary density (red density plot). (E–G) Normalized interchromosomal contact counts between 4-Mb genomic regions grouped by the number of HDS boundaries per region in the different 1,6-HD–treated time points (E), flavopiridol (F), and triptolide (G) treatments. (H) Example of association of nuclear speckles with HDS-boundary cluster vs. unaffected boundary on the same chromosome by immune-DNA FISH. Representative immune-DNA FISH image showing cells stained with the nuclear speckle marker SON (yellow), hybridized with probes targeting an unaffected boundary (red) and an HDS-boundary cluster (green). Inset shows a higher magnification of one cell, showing two alleles of an unaffected (red arrow) and an HDS-boundary cluster (green arrow). (I) Quantitation of immune-DNA FISH images shows that HDS boundary clusters (green) are significantly closer to the nuclear speckle marker (SON) than unaffected boundaries (red). The 1,6-HD treatment increases the median distance between HDS boundary and nuclear speckles, but it has no impact on the non-HDS–boundary association with speckles (n.s. - not significant). (J) Visual interpretation of immune-DNA FISH data.